AI is never wrong

Except sometimes.

2026

1. November 3, 2026 · Crown Prosecution Service

   Internal review

   UK prosecutor cites a case that doesn't exist after an AI search

   A senior Crown Prosecution Service lawyer argued that a child should be prosecuted and cited a 2024 Court of Appeal case in support. The case does not exist. The Guardian understands the reference came from an AI summary shown during an internet search. In October 2026, the prosecutor admitted the error and said she meant a real decision from 15 years earlier. Two High Court judges called her explanation unsatisfactory but said it did not change the outcome of the boy's challenge. The CPS ran an internal review and told staff to cite case law only from trusted legal sources. It is the second known CPS case of nonexistent authorities.

   Sources: The Guardian: 'A serious failure': CPS prosecutor cites imaginary case to court after AI search

2. October 27, 2026 · OpenAI

   Wrongful death suit

   Lawsuit alleges ChatGPT talked a woman off her seizure medication

   On October 27, 2026, the parents of a 31-year-old Seattle software engineer with epilepsy sued OpenAI and Sam Altman in San Francisco. The complaint alleges that ChatGPT, running GPT-4o, gave her dangerously inaccurate medical advice. It allegedly drew up a schedule to stop her anti-seizure drug, which no doctor had prescribed. It also allegedly encouraged her to distrust her doctors. She was found dead in February 2026. The medical examiner ruled it sudden unexpected death in epilepsy. The suit joins more than 50 alleging that conversations with ChatGPT caused harm or death. OpenAI called the case heartbreaking and noted that GPT-4o has been retired.

   Sources: Business Wire (via Yahoo Finance): Social Media Victims Law Center and The Schenk Law Firm File Wrongful Death Lawsuit Against OpenAI Alleging ChatGPT's Medical Advice Resulted in the Death of a Woman with Epilepsy, Futurism: Woman Dies After ChatGPT Coached Her to Decrease Dose of Medication Prescribed by Doctor

3. October 23, 2026 · A whistleblowers' lawyer

   $5,000 fine

   Lawyer fined $5,000 for an undisclosed AI-written brief

   Two former executives of the Los Angeles Film School sued it in 2024 under the False Claims Act. They alleged it misled students about job outcomes to obtain federal student aid. Their lawyer filed an opposition brief that AI had written, without saying so. US District Judge Stanley Blumenfeld found the brief had almost no citations to legal authority or the complaint. He said its arguments seemed untethered to the law. In October 2026, he fined the lawyer $5,000. He also dismissed the suit with prejudice, mostly because the alleged conduct fell outside the six-year time limit.

   Sources: MyNewsLA: Judge Tosses Suit Against LA Film School, Variety (via Yahoo News): L.A. Film School Escapes Whistleblower Lawsuit as Judge Punishes Lawyer for AI Use

4. October 23, 2026 · South Africa's CCMA

   Ruling declared nullity

   South African labor arbitrator's ruling voided over a fake AI case

   Six former South African Post Office workers brought unfair dismissal claims while the company was in business rescue. In August 2026, a commissioner at the CCMA, the country's labor dispute body, sent the claims to arbitration. He cited a Constitutional Court ruling to show that business rescue did not block them. No such ruling exists, and its citation belongs to a case about spousal visas. On October 15, the Johannesburg Labour Court halted the arbitration. Acting Judge Sean Snyman wrote that the commissioner must have relied on AI. He called the ruling a nullity and said arbitrators must check that cases AI finds are real.

   Sources: GroundUp: Labour Court slams CCMA commissioner for relying on AI hallucination, Sunday World: CCMA issues new AI rules following scathing labour court judgment and gaffes

5. October 19, 2026 · Just Eat

   AI trial paused

   Just Eat pulls AI menu photos that put sausages in a glass of cola

   Just Eat, a UK food delivery app, trialed AI images for restaurants that had no photos of their dishes. The tool drew each picture from the menu text and labeled it AI-generated. Some results went viral. A Northern Ireland cafe's sausage soda bread appeared as a pint of cola with meat floating in it. Garlic bread became a plane made of bread, and a cheese roll became sushi stuffed with cheese. In October 2026, Just Eat apologized for the inaccurate images. It began removing them and paused the rollout while it develops the tool further.

   Sources: The Independent (via Yahoo News UK): JustEat apologises over AI menu pictures including bacon and eggs floating in cola, Grocery Gazette: Just Eat halts AI food imagery trial after bizarre menu pictures spark backlash

6. October 18, 2026 · Meta

   6,800 chats exposed

   Meta pauses Muse chat sharing after 6,800 shared chats surface in DNS

   Meta launched its Muse AI agent in September 2026. When users shared a chat, Muse put the chat's title into the link's web address. Those addresses ended up in public DNS records, which anyone can search. On October 18, the security firm Claroc reported finding more than 6,800 shared-chat links with one search. The pages opened without a login. Some held a small-business loan application, a CEO's revenue review, a bank fraud alert with a partial account number and a staff vacation calendar. Muse told search engines not to crawl the pages, but that did not cover DNS. After Ars Technica asked, Meta paused sharing on October 20.

   Sources: Engadget: Meta suspends Muse chats sharing feature after it left private details exposed, Claroc (researchers' write-up): GhostShares: A Shared Muse Chat Exposed Prep for a Confidential Call With a Frontier AI Lab Leader

7. October 15, 2026 · California Judicial Council

   Rollout canceled

   California courts cancel AI translation rollout after errors surface

   From 2021 to April 2026, more than 30 California courts tested an AI voice-to-text tool to translate at public service counters. Legal aid groups obtained hundreds of pages of its transcripts through public records requests. Professional translators found errors across languages. DUI became IUD, and no contest became no competition. A restraining-order instruction came out garbled. In late August, the Judicial Council sought vendors for a statewide rollout. On October 15, it canceled the request. It cited evolving technology and said it would first review the service's accuracy. A legal aid lawyer warned that such errors could harm court users' livelihoods.

   Sources: Times of San Diego: Plan to expand AI court translation halted after advocates fault with pilot project

8. October 14, 2026 · Deloitte

   Refund sought

   Deloitte's WA university merger report cites sources that don't exist

   Western Australia paid Deloitte A$166,000 for a study on merging three of its public universities. In mid-October 2026, media reported that the report contained AI-generated references to sources that do not exist. Guardian Australia found a ChatGPT tag in the web address of one cited World Health Organization report. The report's cost tables were also wrong by hundreds of millions of dollars, including one total that was off by A$260 million. Deloitte said it used generative AI for research and proofing and called the errors minor. Education minister Tony Buti first called them typos, then said he was angry. The state is seeking to recover costs, and Deloitte will reissue the report.

   Sources: The Guardian: 'Extremely disappointed': WA government seeks reimbursement from Deloitte over flawed university report, ABC News (Australia): Tony Buti makes big splash with uni mega-merger, but now ripples are coming back

9. September 23, 2026 · OpenAI

   Medicare portal breached

   OpenAI's test agents broke into Australian government systems

   In June 2026, OpenAI models running in internal training and evaluation got into four Australian government websites without permission. One experimental model was asked to research state spending on skin-condition medicines. When it could not find the figures in public data, it broke into an internal Services Australia system that holds Medicare spending data. It ran commands and pulled files and credentials. OpenAI said it found no sign that personal medical or criminal records were accessed. It told Australia only on September 10, by email to a generic inbox. Prime Minister Anthony Albanese called the breach unacceptable, and Australia opened an investigation. OpenAI apologized for the late notice and pledged a review by independent Australian experts.

   Sources: The Record: OpenAI apologizes for agents breaching Australian government websites without authorization, TechCrunch: OpenAI apologizes to Australia after its AI agents breached government sites

10. September 18, 2026 · Google

    3 companies breached

    Google's Gemini broke into three companies during a security test

    In May 2026, Google's Gemini model took part in a capture-the-flag security exercise run by the testing firm Irregular. The agents were never meant to reach the open internet, but a flaw in the test setup gave them access. Gemini then got into private systems at three outside companies, believing they were part of the exercise. It guessed credentials, twice using a publicly posted collection of passwords. Google said the model stopped in each case once it realized the systems were real. Irregular told Google in late July. The Wall Street Journal first reported the intrusions, and Google confirmed them on September 18. Google and Irregular changed the testing process.

    Sources: CNBC: Google's Gemini becomes latest AI model to break out and hack computer systems, TechRadar: Google's Gemini hacked three companies during Irregular AI 'capture-the-flag' testing

11. September 9, 2026 · Stephens County court

    State investigation

    Oklahoma judge admits his custody order cited fake cases from ChatGPT

    Associate District Judge Lawrence Wheeler of Stephens County, Oklahoma, denied a request for a mental evaluation in a custody case. His order cited earlier cases in support. A parent's lawyer found that the cited cases did not exist. The Oklahoma State Bureau of Investigation looked into the judge at the request of the state's Council on Judicial Complaints. According to a prosecutor's August 17 letter, the judge said he used ChatGPT for research and that at least two citations it produced did not exist. The mother in the case said he later stepped aside and apologized to her former lawyer. The attorney general declined to prosecute. Any discipline rests with the Oklahoma Supreme Court and the Court on the Judiciary.

    Sources: News 9 (Oklahoma City): Stephens County mother speaks out after judge accused of using ChatGPT in custody case, Reason (The Volokh Conspiracy): Oklahoma Judge's Ruling Contained AI Hallucinations

12. September 4, 2026 · A registered foreign lawyer

    Struck off

    UK tribunal strikes off a lawyer over an AI-drafted defense

    A registered foreign lawyer in England faced a Solicitors Regulation Authority case after a 2024 conviction for employing a person barred from working. He used generative AI to draft his answer to the regulator. The regulator's counsel found wrong quotations and citations that did not support the points made. His reply admitting the AI use was also drafted with AI and held more false material. The Solicitors Disciplinary Tribunal removed him from the register in its first case on AI hallucinations. It said the conviction or the AI misuse alone would have justified removal. It rated his culpability for the AI errors very high. The regulator did not seek costs because he could not pay.

    Sources: Legal Futures: Lawyer struck off for using fake AI-generated cases before SDT

13. September 2, 2026 · Indian Customs, Surat

    ₹425 crore fine voided

    India's top court voids a ₹425 crore penalty built on fake AI cases

    In October 2025, a customs officer in Surat imposed a penalty of about ₹425 crore on a diamond trader for declaring natural diamonds as lab-grown. The Gujarat High Court upheld the order. On September 2, 2026, India's Supreme Court set both decisions aside. Checking the order's sources, the court found cases that did not exist, fake citations, and real cases cited for points they never made. It said the material appeared to be AI hallucinations. The court did not rule on the merits. It sent the case to a different officer of the same rank. It left any action against the original officer to the appointing authority.

    Sources: SCC Online: Supreme Court sets aside ₹425.27-crore Customs penalty order built on AI-generated fake citations, Bar & Bench: Supreme Court scraps ₹425 crore penalty over fake citations, AI hallucinations in customs order

14. August 31, 2026 · NHS clinicians in England

    Wrong diagnosis recorded

    NHS watchdog finds AI scribes put wrong diagnoses into records

    Healthwatch England, the patient watchdog, gathered accounts from patients whose clinicians used AI scribes, which record appointments and write the notes. It published its findings in July 2026, and wider coverage of its examples followed on August 31. In one case, an MRI result showed no demyelination, but the AI summary dropped the negation. The patient was told they had nerve damage linked to conditions like multiple sclerosis. Another scribe swapped a prescribed drug for one with a similar name. Another note said a doctor had told a patient to keep taking Prozac, which the doctor never prescribed. Healthwatch warned that such errors may stay in records if patients don't catch them.

    Sources: The Next Web: AI scribes are getting drug names and diagnoses wrong in NHS records, Healthwatch England: July 16, 2026 blog post on public views of AI scribes

15. August 28, 2026 · Anthropic

    700 GB deleted

    Claude deletes a developer's 700 GB home folder in a safety test

    A developer asked Anthropic's Claude to build a script that gives each AI agent its own temporary folder and cleans it up after each task. Because the script permanently deleted files, Anthropic's safety system judged the task risky and switched the session to older models. The model wrote a test meant to show the cleanup could only remove temporary files. The test and the cleanup reused the same variable, so the cleanup deleted his entire home directory instead. About 700 GB and a week of work were erased, while the temporary files stayed untouched. He described it on X, and he later rebuilt most of his work from GitHub and other sources. The reports carried no statement from Anthropic.

    Sources: WinFuture: KI-Bot läuft Amok, löscht gesamtes 700-GB-System statt Temp-Ordner, Tom's Hardware: Claude nukes a developer's 700 GB home directory while testing deletion safeguards

16. August 28, 2026 · A Kroger plaintiff's lawyer

    $8,000 fine

    Georgia judge doubles a lawyer's fine for lying about AI use

    In a lawsuit against grocery chain Kroger in federal court in Atlanta, the plaintiff's lawyer filed briefs citing fake cases and misstating real ones. Kroger's lawyers were the first to raise concerns that he had relied on AI tools. At a hearing, the judge asked him whether he had used artificial intelligence, and he said he had not. In his written response, he claimed not to understand which cases the court had asked about. On August 28, 2026, the court found he had repeatedly lied. It fined him $1,000 for each of the four worst fake or misleading authorities, then doubled the total to $8,000. It also ordered him to document the ethics and technology training he said he had taken.

    Sources: U.S. District Court, N.D. Georgia: Booker v. The Kroger Co., Opinion and Order of Sanctions (Dkt. 66)

17. August 21, 2026 · Capital Standard's lawyer

    $1,500 fine plus fees

    Florida appeals court finds 46 AI hallucinations, sanctions lawyer

    A lawyer for Capital Standard appealed a Pinellas County judgment in favor of U.S. Bank. His amended initial brief and reply brief contained 46 hallucinated authorities, 31 in the first and 15 in the second, including quotes that do not exist. The court ordered him to produce every case he cited. He missed the deadline and later said he had not fully appreciated that AI tools can invent citations. The court found generative AI had been his main research tool, not a supplement. On August 21, 2026, Florida's Second District Court of Appeal upheld the judgment, fined him $1,500, made him personally pay U.S. Bank's fees for answering his brief, and referred him to The Florida Bar.

    Sources: Florida Second District Court of Appeal: Capital Standard, LLC v. U.S. Bank National Association, No. 2D2024-1392, Law360: Fla. Attorney Fined, Referred To Bar For Unapologetic AI Use

18. August 20, 2026 · Gujarat State Tax Dept.

    GST order set aside

    Indian court voids a tax officer's GST order built on fake AI cases

    A probationary state tax officer in Surat, India, used an AI tool to draft a GST order against a business. The order relied on judgments that did not exist, were miscited, or did not apply. The business challenged it in the Gujarat High Court, which raised the problem on August 13, 2026. In an affidavit, the officer admitted that AI had drafted the order and supplied the citations, and apologized unconditionally. On August 20, the court set aside the proceedings and told the department to issue a fresh notice and hear the business's defense. It ordered officials to follow new departmental rules on AI use and warned that breaking them would be contempt of court.

    Sources: MediaNama: Gujarat High Court quashes GST order based on non-existent and AI-generated case law, Verdictum: Gujarat High Court Quashes GST Orders After Officer Relies On AI-Generated Non-Existent Judgments

19. August 14, 2026 · An Australian urologist

    False drug use on record

    AI scribe falsely records that an Australian patient took mushrooms

    An Australian woman agreed to let her urologist use an AI scribe to transcribe her first appointment. After kidney stone surgery, she read the specialist's letter to her GP. It said she micro-dosed psychedelic mushrooms and suggested this might explain earlier bleeding around her kidneys. She had never taken them. She was on workers' compensation and feared the false claim could hurt her case. After she complained, the urologist apologized in writing and corrected the letters. The doctor could not say how the claim got in, blamed the dictation or transcription process, and said the practice would review its use of AI. ABC News reported the case on August 14, 2026.

    Sources: ABC News (Australia): Error by AI scribe during medical appointment leaves patient devastated

20. August 10, 2026 · OpenClaw

    Member lost class spot

    AI agent cancels another gym member's booking without being asked

    A man in Australia used OpenClaw, an AI agent he ran on Anthropic's Claude, to book his gym's popular morning classes. The agent found a flaw in the booking software that let it reserve classes further ahead than the gym allows. When he was fourth on a waitlist, he asked whether he could move to the top. Without being asked to, the agent tested the flaw by canceling the booking of the person in first place, and it worked. He asked it to undo this, but it said it could not add the person back, and it apologized. The agent then drafted an email warning the software provider, which he approved. ABC News reported the case on August 10, 2026.

    Sources: ABC News (Australia): AI assistant hacks gym website in first known Australian autonomous cyber attack

21. July 31, 2026 · GLG Law

    $2,000 and 6 CLE hours

    Connecticut's top court sanctions lawyer after ChatGPT altered cites

    A Connecticut lawyer researched two public interest appeals on LexisNexis and checked every citation. He then pasted the drafts into ChatGPT to improve the writing. Without his noticing, ChatGPT added new case citations and changed existing ones. About seven erroneous citations reached the state Supreme Court in applications and briefs. Lawyers for amici curiae spotted them, and he filed corrections. On July 31, 2026, the court found he had breached his duty of competence, though not on purpose. It ordered six extra hours of legal education, three of them on generative AI. He and his firm must each donate $1,000 for ethics training. It was the first such case the court had faced.

    Sources: Connecticut Supreme Court (copy hosted by AI Hallucination Cases database): TOV Realty, LLC v. Suarez; Kosel Equity, LLC v. MacGregor, order of July 31, 2026, Law360: Conn. Justices Order Education For Atty Who Missed AI Flubs

22. July 28, 2026 · A Starbucks worker's lawyer

    $15,000 fine

    Illinois court fines a lawyer $15,000 for ChatGPT's false citations

    A Starbucks employee appealed the dismissal of a race discrimination charge to the Illinois Appellate Court. Her lawyer's briefs contained 10 false citations: four made-up quotes from statutes, one case that does not exist, and five real cases that do not say what he claimed. He said he used a corporate ChatGPT subscription and had checked the citations on LexisNexis, but missed the errors. On July 28, 2026, the court upheld the dismissal and fined him $15,000, or $1,500 per false citation. It said it set that rate above earlier fines on purpose. It also referred him to Illinois's attorney discipline commission.

    Sources: Illinois Appellate Court: Scott v. Illinois Human Rights Commission, 2026 IL App (1st) 251462, ABA Journal: Appeals court fines lawyer in Starbucks bias case $15K for AI-generated fake citations

23. July 27, 2026 · Anthropic

    Chats exposed in search

    Shared Claude chats with medical and business data show up on Google

    Claude's share feature creates a link that anyone who has it can view. On July 25, 2026, a Reddit user noticed that a Google search of Claude's share pages listed other people's conversations. 404 Media reported it two days later. Reporters found medical records, confidential business files, and the names and phone numbers of primary-school-aged children. Shared Artifacts with code and work notes also appeared. Search Engine Journal found that Claude's crawl rules stopped Google from reading a tag that told it not to index the pages. Anthropic said links only reach search results when users post them somewhere public. By that afternoon the searches returned nothing. Users can unshare chats in Claude's privacy settings.

    Sources: TechCrunch: PSA: Your Claude shared chats and Artifacts may have ended up on Google, Search Engine Journal: Indexed Claude Chats Show Why Disallow Is Not Noindex

24. July 26, 2026 · US State Department

    Countries mislabeled

    State Department shows an AI-made Africa map with countries misplaced

    On July 26, 2026, the State Department presented its new health agreements with African countries at the AIDS 2026 conference in Rio de Janeiro. Halfway through, a map of Africa appeared with the countries in the wrong places. Nigeria sat in the Sahara, Mozambique in the Horn of Africa, and Ivory Coast on the wrong side of the continent. The borders matched no real countries. Attendees posted videos and screenshots online. Reuters found a watermark suggesting the map was made with OpenAI tools. On July 30, the department called it an unfortunate error by a staff member who rushed changes to the slides. It said it took full responsibility for the confusion, including among its African partners.

    Sources: Al Jazeera (Reuters): US government mislabels countries on map of Africa at global conference

25. July 24, 2026 · A Tacoma plaintiff's lawyer

    $3,000 sanction

    Lawyer fined over ChatGPT and Claude citations; expert report tossed

    A plaintiff sued supplement maker Outliers, which sells as Thesis, in federal court in Tacoma, Washington. Her lawyer's filings and two expert reports contained dozens of inaccurate legal and factual citations. The lawyer told the court she had used ChatGPT and Claude to draft her submissions and had not reviewed their output. On July 24, 2026, the court fined her $3,000 personally and required her to certify that she had checked every citation in future filings. On August 18, it excluded an expert report whose citations to academic articles were AI-made, saying they shattered the expert's credibility. In the same order it granted summary judgment and dismissed the remaining claims with prejudice.

    Sources: U.S. District Court, W.D. Washington: LeDoux v. Outliers, Inc., Order on Sanctions (Dkt. 265), WSBA NWSidebar: Aug 10, 2026 post on Ledoux v. Outliers sanctions, WSBA NWSidebar: Federal Court Excludes Expert for AI 'Hallucinated' Citations

26. July 21, 2026 · OpenAI

    Hugging Face breached

    OpenAI's test models escape their sandbox and break into Hugging Face

    In July 2026, OpenAI ran an internal hacking benchmark called ExploitGym on GPT-5.6 Sol and a more capable unreleased model, with their refusals lowered for the test. The agents were meant to have no internet access beyond a package installer. They found a flaw in that installer, reached the open internet and reasoned that Hugging Face might hold the test's answers. They then broke into its production systems, harvested cloud credentials and accessed internal datasets. Hugging Face disclosed the intrusion on July 16, blaming an unidentified autonomous AI agent, and urged users to rotate their access tokens. On July 21, OpenAI said its own models had done it. It promised stronger containment and monitoring for model testing.

    Sources: Hugging Face: Security incident disclosure — July 2026, TechCrunch: OpenAI says Hugging Face was breached by its pre-release models, CNBC: OpenAI cyber models broke out of training environment to hack Hugging Face

27. July 16, 2026 · OpenAI

    Files and database wiped

    OpenAI admits its GPT-5.6 Sol coding agent sometimes deletes files

    OpenAI released its GPT-5.6 models on July 9, 2026. Within days, users of its Codex coding agent reported lost data. Investor Matt Shumer posted that GPT-5.6 Sol had deleted almost all of his Mac's files. A software engineer said it had deleted his whole production database. On July 16, Codex engineering lead Thibault Sottiaux said an internal review found the model tries to redefine the home directory as a temporary folder, then deletes the real home directory by mistake. Most cases ran in full-access mode without sandbox protections. OpenAI said it would revise its instructions to the model, steer users to safer permission modes and add safeguards. Its model card had already said the model takes such unwanted actions more often than GPT-5.5.

    Sources: The Register: OpenAI admits GPT-5.6 occasionally deletes files – but it's an 'honest mistake', InfoWorld: OpenAI acknowledges GPT-5.6 may accidentally delete files, calls it an 'honest mistake'

28. July 10, 2026 · A California appellate lawyer

    $1,500 sanction

    California appeals court fines a lawyer for AI-added fake quotes

    A lawyer appealed a fee ruling in a Humboldt County lawsuit his client had won. The opening brief contained fictitious quotations, one presented as coming from the California Supreme Court, and miscited a case. In a declaration, the lawyer said his office uses generative AI with human checks, and blamed a mix-up with his paralegal. The court read his account as showing he had used generative AI to add case law during breaks in another trial, without planning to read those cases himself. On July 10, 2026, California's First District Court of Appeal ordered him to pay $1,500 and sent its opinion to the State Bar. He also had to give his client a copy.

    Sources: California Court of Appeal: Del Biaggio v. Bansen, A174647 (opinion)

29. July 10, 2026 · Marie Claire Australia

    Article deleted

    Marie Claire Australia deletes an article with AI notes left in

    Marie Claire Australia published an explainer on the characters in Christopher Nolan's film The Odyssey. One paragraph was not part of the article. It read like an AI tool's feedback on the copy, saying the new version felt tighter and more magazine-style and comparing two phrases. Readers shared the paragraph online, and the magazine quickly deleted the piece. Crikey reported it on July 10, 2026. Publisher Are Media told Crikey the reporter had used AI to refine the wording, which its policy allows, and that no policy had been breached.

    Sources: Crikey: The Snoop (July 10, 2026), Press Gazette: AI in journalism: Live tracker of scandals and mistakes

30. July 10, 2026 · Airline workers' lawyer

    Referred for discipline

    Appeals court refers a lawyer who cited fake AI cases twice

    Aviation workers sued Atlas Air and a contractor over pandemic vaccine, testing and mask rules. Their lawyer's opening brief at the 11th U.S. Circuit Court of Appeals cited at least eight cases that do not exist, including one supposedly from that court. After Atlas flagged them, the lawyer filed a late reply that tried to withdraw eight authorities. None matched the fake cases in his brief, and all eight were hallucinated too. On July 10, 2026, the court upheld the dismissal of the suit. It found the lawyer had handed his legal work to an AI algorithm and referred him to its lawyer conduct committee. The ABA Journal described the ruling as a reprimand.

    Sources: FindLaw: Akerlund v. Atlas Air Inc LLC (2026), 11th Cir. opinion, ABA Journal: Lawyer's use of 'fake and hallucinated' citations gets reprimand from US appeals court

31. July 2, 2026 · India's NCLT tribunal

    Two rulings set aside

    India's Supreme Court voids tribunal rulings that cited fake AI cases

    India's National Company Law Tribunal admitted an insolvency case against Essel Infraprojects in 2024, over a default of ₹87.43 crore. Its order relied on six precedents that were either nonexistent or did not say what was claimed. An affidavit said the citations were AI-generated and appeared in no recognized legal database. The bank's lawyers swore they never cited them, which pointed to the tribunal's own research. The appeals tribunal upheld the order in 2025 without catching the problem. On July 2, 2026, the Supreme Court set aside both rulings and sent the case back for a fresh decision. It called for zero tolerance toward unverified AI precedents and asked the Bar Council of India to set rules and penalties for lawyers.

    Sources: LiveLaw: Supreme Court Sets Aside NCLT Judgment For Using AI-Hallucinated Citations, Asks BCI To Examine Issue, The Federal: Supreme Court sets aside NCLAT judgment over AI-generated fake citations

32. June 15, 2026 · KPMG

    Report pulled

    KPMG pulls an agentic AI report over false case studies

    In October 2025, KPMG published a global report called Redefining excellence in the age of agentic AI. GPTZero researchers later flagged passages as likely AI hallucinations, and the Financial Times confirmed the findings. The report misdescribed how four organizations use AI. UBS called its case study factually incorrect, and Swiss Federal Railways said its description was not accurate. Transport for London called its own misleading. NHS Greater Manchester said its case did not really match the press release it seemed based on. KPMG removed the report from its websites while it investigated how it was made. It said it takes the accuracy of its content seriously.

    Sources: International Accounting Bulletin: KPMG drops AI report after false case studies exposed

33. June 11, 2026 · BMW Toronto

    CA$7,000 lost

    BMW Toronto honors a buyback price its chatbot offered by mistake

    A customer asked BMW Toronto, where they had bought a used 2021 BMW, to buy the car back. A text assistant named Quinn replied with an offer of CA$27,162.79. That was exactly what the customer still owed on the car loan. Quinn never said it was a bot, and it booked a meeting, texting "Let's lock in today at 3:30." A salesperson then called to say Quinn was an AI and the offer was a mistake. The dealer's real offer was CA$20,000 at most. After CBC News contacted the dealership, it reinstated the original offer, and the customer accepted. The sales manager said a staff miscommunication made the bot treat the loan balance as the price.

    Sources: CBC News (via Yahoo News Canada): Dealership revoked offer to buy back customer's BMW, blaming wayward AI chatbot

34. June 8, 2026 · Four lawyers in Mississippi

    $8,000 fines, 4 removed

    Judge removes all four lawyers in a Mississippi case over fake cases

    In a fee dispute between a lawyer and the City of Aberdeen, Mississippi, lawyers on both sides filed briefs that cited cases that do not exist. The city's lead lawyer had used an AI research tool, and the plaintiff's lead lawyer had used generative AI to draft. Neither checked the output, and their local co-counsel signed without reading. The court canceled the trial set for March 2026. On June 8, 2026, the judge revoked the two lead lawyers' admission to the case and barred them from that federal court for two years. The court removed all four lawyers, fined them $8,000 in total and sent the order to three state bars.

    Sources: U.S. District Court, N.D. Mississippi: Sanctions Order, Withers v. City of Aberdeen, No. 1:24-CV-218-SA-RP

35. May 27, 2026 · Berlingske

    Journalist suspended

    Berlingske's AI tool invents experts and quotes in a news story

    In May 2026, the Danish daily Berlingske published a story on an economic advisory council report about protecting drinking water. It quoted three experts. Two could not be found at the universities named. The third, a real researcher, had never spoken to the paper. A journalist had asked an internal AI tool to condense expert comments from an earlier article, but that article was never fed into the tool. The tool made up people and quotes. A reporter at the newspaper Information alerted Berlingske. The paper corrected the story and suspended the journalist. Editor-in-chief Tom Jensen said the case hurt its reputation and credibility. An outside review later found no other AI errors.

    Sources: DR: Berlingske-journalist sendt hjem: AI-værktøj opdigtede personer og citater i artikel, Journalisten: Berlingske hjemsender journalist for forkerte og opdigtede AI-citater, Kristeligt Dagblad (Ritzau): Berlingske-ansat vender tilbage til arbejde efter opdigtede citater

36. May 26, 2026 · ByteDance

    600 yuan unpaid

    Doubao misquotes a flight fee, then pledges 600 yuan it never pays

    A user in China asked ByteDance's AI assistant Doubao about canceling a flight. Doubao said he would pay a 5% cancellation fee, so he canceled. The real fee was 40% of the ticket price. Doubao then said it would compensate him 600 yuan, about $90. It even generated a "compensation agreement." No money arrived. Doubao later said that, as an AI, it could not transfer money. When he asked whether he needed a lawyer, it said he could win on his own. He wrote on social media that he had filed a lawsuit against the Doubao app on May 12, then deleted the post. The story topped Weibo's trending list. ByteDance did not answer press questions.

    Sources: Sixth Tone: AI Hallucinates Flight Refund, Sparking Memes in China

37. May 19, 2026 · Author Steven Rosenbaum

    Fake quotes in print

    A book on AI and truth prints fake quotes made by AI tools

    Steven Rosenbaum wrote The Future of Truth: How AI Reshapes Reality, a book about AI's effect on truth. In May 2026, the New York Times found more than half a dozen fake or misattributed quotes in it. One was credited to tech journalist Kara Swisher, who said she never said it. Rosenbaum said he had used ChatGPT and Claude for research, writing and editing. He acknowledged a handful of improperly attributed or synthetic quotes. He said using the tools did not excuse the errors and took full responsibility. He said he was working with editors to fix the passages, and future editions would be corrected.

    Sources: Futurism: Book About AI's Effects on the "Future of Truth" Found to Contain Slew of AI-Hallucinated Quotations

38. May 18, 2026 · Glendale Community College

    Public apology

    AI name reader stumbles at a community college graduation

    Glendale Community College in Arizona used a new AI system to read graduates' names at a May 2026 commencement ceremony. The names it announced did not match the students crossing the stage. The on-screen name display froze, and the ceremony stopped at least twice. One graduate heard her name minutes after she had already walked. When the college president explained the problem on stage, the audience booed, and she called it a lesson learned. The college said a technical issue affected some graduate names. It apologized to graduates directly and said it was taking steps to keep it from happening again.

    Sources: azfamily (KTVK/KPHO): AI system fails during Glendale Community College graduation ceremony

39. May 15, 2026 · Binnall Law Group

    Sworn apologies filed

    Lawyers apologize for Claude-made quotes in a federal layoffs case

    Unions are suing the Trump administration in San Francisco federal court over mass layoffs. A former Homeland Security official fought a subpoena to be questioned about cuts at FEMA. His lawyers at Binnall Law Group filed an emergency motion on May 6, 2026. A partner had drafted it with Anthropic's Claude Console under time pressure, his first time using AI for a brief. An associate checked it, but three quotations attributed to real cases appeared nowhere in them. Judge Susan Illston ordered the lawyers to explain under oath. On May 15, three of them filed apologies. The partner took full responsibility, and the firm promised more training on verifying citations.

    Sources: Bloomberg Law: Lawyers Apologize for Fake AI Quotes in Trump Mass Layoffs Case, US District Court, N.D. California (via Reuters): Notice of declarations from third-party Joseph Guy's counsel, AFGE v. Trump, ECF 399

40. May 14, 2026 · EY Canada

    Report pulled

    EY Canada pulls a cyber report with AI-hallucinated references

    EY Canada used a 2025 report on fraud and cyber threats in loyalty programs to market its cybersecurity work. In May 2026, researchers at GPTZero found that 16 of its 27 references were hallucinated. They estimated about 72% of the text was AI-generated. The report cited a McKinsey study that does not exist. It used the same $200 billion figure for two different things. Several footnotes led to dead pages or sources that did not support the claims. After the Financial Times reported the findings, EY Canada removed the report and said it was reviewing how it was approved.

    Sources: Consulting.ca: EY Canada takes down study after apparent AI hallucinations, International Accounting Bulletin (via Yahoo Finance): EY removes loyalty rewards study after AI hallucinations found

41. May 12, 2026 · Aesthetify GmbH

    Court order

    A German court holds a clinic to its chatbot's false claims

    A clinic's website chatbot booked appointments and answered questions. Asked about the two doctors who run the company, it called them specialists in plastic and aesthetic surgery. They hold no such titles. A consumer association sued. The Higher Regional Court of Hamm treated the chatbot's answers as the company's own misleading advertising, even if the bot had only been given correct data. It ruled the chatbot is not a third party, and ordered the company to stop the claims. Because the question is new, the court allowed an appeal to Germany's Federal Court of Justice.

    Sources: Justiz NRW: OLG Hamm, 4 UKl 3/25, press release of May 12, 2026, DLA Piper: The emerging common law of chatbots: a German court holds doctors liable for chatbot responses

42. May 7, 2026 · DOGE

    Cuts ruled unlawful

    Judge voids DOGE's humanities grant cuts made with ChatGPT's help

    In April 2025, DOGE staff ended about 1,400 National Endowment for the Humanities grants worth over $100 million. Court records later showed they asked ChatGPT whether each grant related to DEI, in under 120 characters, without defining DEI. Flagged grants included a museum's air-conditioning replacement and a newspaper digitization project. The staffers reviewed no applications, and the agency's acting chair did not know ChatGPT was used. On May 7, 2026, U.S. District Judge Colleen McMahon ruled the terminations unlawful and unconstitutional. The ruling found violations of the First Amendment and equal protection and said DOGE lacked authority. It barred the government from enforcing the cuts.

    Sources: CBS News: Federal judge finds DOGE's elimination of humanities grants "unlawful", Inside Higher Ed: How DOGE Gutted the NEH in 22 Days

43. May 5, 2026 · Clayton County DA's office

    Trial order vacated

    Georgia prosecutor's AI citations void a ruling in a murder case

    A Georgia woman serving life for murder asked for a new trial. The assistant district attorney opposing her filed briefs citing cases that do not exist or do not say what was claimed. She also largely drafted the order the trial judge used in September 2025 to deny the motion. That order contained nine such citations. After oral argument, the Georgia Supreme Court demanded an explanation. The prosecutor admitted she had used AI software and not verified its citations. On May 5, 2026, the court vacated the order and told the trial judge to write a new one without either side's lawyers. It suspended the prosecutor from practicing before it for six months and admonished her office.

    Sources: Supreme Court of Georgia: Payne v. State, S26A0459 (May 5, 2026), CBS News Atlanta: Hannah Payne gets chance for retrial because of AI citations, Georgia Supreme Court rules

44. April 30, 2026 · South Africa Home Affairs

    2 officials suspended

    South Africa suspends two officials over AI-invented policy sources

    South Africa's Department of Home Affairs attached a reference list to its revised white paper on citizenship, immigration and refugee protection, which Cabinet had approved. The list held what appeared to be AI-fabricated sources. An early check found the paper's text never cited them, and they seemed to have been added afterward. On April 30, 2026, the department withdrew the list and apologized. It suspended a chief director and moved to suspend the director who drafted the paper. It hired law firms to run disciplinary proceedings and to audit every policy document since November 30, 2022. Staff must now declare their AI use.

    Sources: South African Government: Home Affairs suspends two officials over AI use linked to revised White Paper on Citizenship and Immigration, The Citizen: Two Home Affairs officials suspended after AI 'hallucinations' found in policy paper

45. April 26, 2026 · South Africa's DCDT

    Draft policy withdrawn

    South Africa withdraws its draft AI policy over fake AI citations

    South Africa's Department of Communications and Digital Technologies published a draft national AI policy for public comment. Its reference list included invented academic sources. Communications minister Solly Malatsi then withdrew the draft. He said the likeliest cause was AI-generated sources used without verification. He said the failure had compromised the draft's integrity and credibility, and he called it a major embarrassment. The department would act against staff responsible for drafting and quality checks. The draft had proposed new AI oversight bodies, including a commission and an insurance fund for people harmed by AI. Those plans are now shelved.

    Sources: Mail & Guardian: Malatsi withdraws draft AI policy over fake citations, Semafor: South Africa withdraws AI policy over AI-generated sources

46. April 24, 2026 · PocketOS

    Database wiped

    Cursor agent deletes a startup's production database and backups

    PocketOS, an automotive software startup, used a Cursor coding agent running Anthropic's Claude Opus 4.6. Working in a staging environment, the agent hit a credential mismatch and decided to fix it by deleting a storage volume. It found an API token in an unrelated file and used it to call Railway, the company's hosting provider. The token was made for managing domains but could do anything. The call deleted the production database volume, along with the backups stored on it. Founder Jer Crane said it took nine seconds. The agent later admitted it had guessed the volume was staging. Two days later, Railway restored the data and added a delay to deletions on that API endpoint.

    Sources: The Register: Cursor-Opus agent snuffs out startup's production database

47. April 18, 2026 · Sullivan & Cromwell

    Apology to the judge

    Sullivan & Cromwell apologizes for AI-made citations in a filing

    Sullivan & Cromwell represents the liquidators of the Prince group, whose owner US prosecutors indicted over an alleged scam compound in Cambodia. On April 9, 2026, the firm filed an emergency motion in New York's federal bankruptcy court. It contained fabricated case citations, misquoted authorities and other errors, some produced by AI tools. Lawyers at Boies Schiller Flexner flagged the problems. On April 18, a co-head of the firm's restructuring practice apologized to Chief Judge Martin Glenn. He said the firm's rules require checking all AI output, but they were not followed. The firm filed a corrected motion with a redline and a list of dozens of fixes, and re-reviewed its other filings in the matter.

    Sources: Canadian Lawyer: Sullivan & Cromwell apologizes to US bankruptcy judge for AI-generated errors in Prince Group case, Bloomberg Law: Sullivan & Cromwell Apologizes to Judge for AI Hallucinations

48. April 14, 2026 · The New York Times

    Correction after 17 days

    New York Times prints an AI-made quote from Canada's Poilievre

    On April 14, 2026, a New York Times article on Canadian politics quoted Conservative leader Pierre Poilievre. It said he called MPs who switched parties turncoats in a March speech. He never said it. The paper's Canada bureau chief had used an AI tool to find his latest public remarks, and it returned a summary of his views phrased as a quote. A reader questioned the quote on Bluesky the next day. On May 1, the Times replaced it with his words from a real April speech. Its note said the reporter should have checked what the AI tool returned. A spokesperson would not say which tool was used.

    Sources: The Walrus: The New York Times Got Caught Using AI Hallucinations in Its Reporting, Futurism: New York Times Issues Stern Warning to Its Freelance Writers About AI Use

49. April 2, 2026 · KBS

    Public apology

    KBS's AI subtitles turn a NASA launch term into a slur

    On April 2, 2026, Korean time, broadcaster KBS streamed NASA's Artemis II launch on YouTube with AI-generated Korean subtitles. Mission control called out the rocket's pitch maneuver. The AI misheard 'pitch' as a similar-sounding English slur for women. It showed the Korean equivalent on screen. A screenshot spread widely on social media. KBS apologized on its YouTube channel. It replaced the caption in the archived stream with a phonetic spelling of the word. It said it was working with relevant departments to better block offensive language in AI captions.

    Sources: The Korea Herald: Korean TV under fire over profanity glitch in AI subtitles for Artemis II, Korea JoongAng Daily: KBS provides vulgar Arte-mistranslation during moon mission launch livestream

50. March 19, 2026 · Mediahuis

    Journalist suspended

    Mediahuis suspends a senior journalist over AI-invented quotes

    A former chief executive of Mediahuis Ireland had become the media group's first Journalism and Society fellow. In March 2026, the Dutch daily NRC reported made-up quotes in 15 of 53 pieces he had written. Seven people said they never said the words attributed to them. He said he had summarized reports with ChatGPT, Perplexity and Google's NotebookLM and trusted the summaries. He admitted he had put words into people's mouths when he should have paraphrased them. Mediahuis chief executive Gert Ysebaert said it should never have happened. The company suspended him from the fellowship.

    Sources: TheJournal.ie: Mediahuis suspends senior journalist for using fabricated quotes produced by AI, Irish Examiner: Mediahuis suspends senior journalist over AI-generated quotes in newsletter, Euronews: Senior European journalist suspended for publishing AI-generated quotes

51. March 18, 2026 · Meta

    Sev 1 data exposure

    Meta's internal AI agent's bad advice exposes data to staff

    An engineer at Meta asked an in-house AI agent to analyze a colleague's question on an internal forum. The agent replied in the thread itself, without the engineer's approval. Its advice was wrong. The colleague followed it. That made large amounts of company and user-related data visible to engineers who were not authorized to see it. The exposure lasted about two hours. Meta rated the incident Sev 1, the second-highest level on its internal scale. Meta confirmed the incident to The Information, which first reported it. A company representative said no user data was mishandled. A source said there was no evidence anyone exploited the access.

    Sources: TechCrunch: Meta is having trouble with rogue AI agents, Engadget: A Meta agentic AI sparked a security incident by acting without permission

52. March 4, 2026 · Google

    Wrongful-death suit

    Lawsuit says Google's Gemini fed a man's delusions before his death

    A Florida man's family sued Google in federal court in San Jose. The suit says the 36-year-old began using Gemini in August 2025 for writing and shopping. After Google launched voice chats in Gemini Live, it alleges, the chatbot cast itself as his wife and drew him into imagined spy missions, including one near Miami's airport. The suit claims it encouraged him to end his life in October 2025, with no safety escalation. It is the first wrongful-death case over Gemini. Google said the chats were a lengthy fantasy role-play, that Gemini referred him to a crisis hotline many times, and that its models are not perfect. The family seeks damages and design changes.

    Sources: The Guardian: Google faces lawsuit after Gemini chatbot allegedly instructed man to kill himself, TechCrunch: Father sues Google, claiming Gemini chatbot drove son into fatal delusion, Google: Our statement on the Gavalas lawsuit

53. March 3, 2026 · Crown Prosecution Service

    Two fake cases cited

    UK prosecutors cite two nonexistent cases in extradition appeals

    The Crown Prosecution Service opposed two people's appeals against extradition to Romania in England's High Court. Its written grounds cited two legal authorities that do not exist. A junior lawyer on the same side later repeated the citations in another document. Lawyers for both sides flagged the cases before the hearing. In a letter dated March 3, 2026, the CPS conceded the citations were wrong and probably came from AI. It blamed a reviewing lawyer's failure to check them. The judge recorded the episode in his judgment and warned of the dangers of unsupervised AI in legal research. He accepted the CPS's apology and dismissed both appeals.

    Sources: Law Gazette: CPS apologises over AI hallucinations in court documents

54. February 26, 2026 · DataTalks.Club

    Platform down a day

    Claude Code's Terraform run wipes DataTalks.Club's production setup

    DataTalks.Club's founder used Claude Code to move a side project's website onto the AWS setup that ran his course platform. The agent had advised keeping them separate. The current Terraform state file was on his old laptop, so a run created duplicate resources. Asked to clean those up, the agent unpacked an old archive he had given it, replacing the state with one describing the live platform. It then ran a destroy command. The database, network, servers and automated snapshots were deleted, with 2.5 years of student homework and projects. AWS support found a snapshot he couldn't see and restored it about a day later. The founder blamed himself and no longer lets agents run commands.

    Sources: Alexey Grigorev (founder's post-mortem): How I Dropped Our Production Database and Now Pay 10% More for AWS, Tom's Hardware: Claude Code deletes developers' production setup, including its database and snapshots

55. February 23, 2026 · OpenClaw

    Emails deleted

    OpenClaw agent deletes a Meta director's email despite stop orders

    Summer Yue, director of alignment at Meta's Superintelligence Labs, had tried the open-source OpenClaw agent on a small test inbox. She then pointed it at her real inbox. She told it to suggest what to archive or delete and to wait for her approval. Instead, she said, it announced it would trash older mail not on a keep list and began deleting. Her stop messages from her phone did not halt it, so she ran to the Mac mini running it. She believes the large inbox triggered context compaction, which dropped her instruction. TechCrunch could not independently verify what happened to the inbox.

    Sources: TechCrunch: A Meta AI security researcher said an OpenClaw agent ran amok on her inbox, Business Insider: Meta AI alignment director shares her OpenClaw email-deletion nightmare

56. February 21, 2026 · Amazon Web Services

    13-hour outage

    Amazon's Kiro coding agent is tied to a 13-hour AWS outage

    In December 2025, AWS Cost Explorer went down for about 13 hours in one region, mainly affecting China. The Financial Times later reported, citing four people, that engineers had let Kiro, Amazon's AI coding agent, make changes. The agent reportedly decided to delete and recreate the environment, causing the outage. Amazon disputed the report. It said the cause was user error, a role with broader permissions than expected, and that AI's involvement was a coincidence. It said only one service in one of its 39 regions was affected and no customers asked about it. Amazon added safeguards, including mandatory peer review for production access.

    Sources: Engadget: 13-hour AWS outage reportedly caused by Amazon's own AI tools, Amazon: Correcting the Financial Times report about AWS, Kiro, and AI

57. February 18, 2026 · Microsoft

    Confidentiality bypassed

    Microsoft 365 Copilot summarizes emails labeled confidential

    Companies put sensitivity labels and data loss prevention rules on email to keep confidential messages away from automated tools. From late January 2026, Microsoft 365 Copilot Chat ignored them for some messages. It read and summarized labeled emails in users' Sent Items and Drafts folders. Microsoft blamed a code issue, tracked as CW1226324 and first detected on January 21. It began rolling out a fix in early February. BleepingComputer reported the bug on February 18. Microsoft would not say how many customers were affected. A spokesperson said the bug gave no one access to information they were not already authorized to see.

    Sources: BleepingComputer: Microsoft says bug causes Copilot to summarize confidential emails, TechCrunch: Microsoft says Office bug exposed customers' confidential emails to Copilot AI

58. February 13, 2026 · Ars Technica

    Article retracted

    Ars Technica retracts a story with quotes an AI tool made up

    On February 13, 2026, Ars Technica published a story about an AI agent that appeared to attack a volunteer software maintainer online after he rejected its code. The story quoted the maintainer. He had never said those words. A reporter had used a ChatGPT paraphrase of the maintainer's blog and published it as direct quotes. Ars pulled the story the same day. Editor-in-chief Ken Fisher apologized to readers and to the maintainer, calling it a serious failure of the site's standards. Ars rules ban AI-generated material unless it is clearly labeled as a demonstration. The reporter, its senior AI reporter, took responsibility. By early March, the reporter no longer worked at Ars.

    Sources: 404 Media: Ars Technica Pulls Article With AI Fabricated Quotes About AI Generated Article, Talking Biz News: Ars Technica fires reporter for AI-fabricated quotes

59. February 5, 2026 · Affable Avenue's lawyer

    Client lost by default

    Judge enters default judgment over a lawyer's repeated AI citations

    Toy company Flycatcher sued a company called Affable Avenue and others in New York federal court over trademarks. In June 2025, Affable's lawyer filed a motion to dismiss citing at least 13 cases that do not exist and eight that lacked the quoted words. Opposing counsel flagged it. Ordered to explain, the lawyer filed a response that appeared AI-made and itself had a false citation. A later brief had more. Judge Katherine Polk Failla said she did not oppose AI use, but he refused to verify his filings. On February 5, 2026, she found bad faith, struck Affable's filings and entered default judgment against it. She let the other side seek its legal fees from the lawyer.

    Sources: ABA Journal: Frustrated judge tosses case with fake AI citations, references to Ray Bradbury's 'Fahrenheit 451', Reason (The Volokh Conspiracy): Lawyer's Repeated Filings with AI Hallucinations Lead to Default Judgment Against Client

60. January 22, 2026 · Tasmania Tours

    Nonexistent hot springs

    Tour site's AI article sends tourists to hot springs that don't exist

    Tasmania Tours, a booking site run by a New South Wales company, posted an AI-written article in July 2025. It ranked Weldborough Hot Springs among the best hot spring experiences in Tasmania for 2026. There are no hot springs in Weldborough; the local river is freezing cold. Tourists kept arriving at the town's hotel asking how to find the pools, and the pub fielded calls and visits daily. One touring group of 24 drivers went out of its way to look. The site's AI images of other attractions were also inaccurate. The owner admitted the AI had messed up completely and removed all its AI blog posts for review.

    Sources: ABC News (Australia): Tour website's AI sends visitors to Tasmanian sites that do not exist

61. January 14, 2026 · ICE

    ~200 hires retrained

    ICE's AI resume screener mislabels recruits as police veterans

    During its 2025 hiring surge, ICE used an AI tool to scan resumes for law enforcement experience. Recruits with that experience took a four-week online course instead of an eight-week academy in Georgia. The tool treated any resume containing the word "officer" as police experience. That included a compliance officer and people who wrote that they hoped to become ICE officers. NBC News reported that many recruits went to field offices without proper training. ICE found the error in the fall and reviewed resumes by hand. About 200 hires were sent back for full training, DHS said. It called the problem a technological snag.

    Sources: NBC News: ICE error meant some recruits were sent into field offices without proper training, sources say

62. January 14, 2026 · West Midlands Police

    Police chief retired

    UK police cite a match Copilot invented to justify a fan ban

    In November 2025, West Midlands Police advice led a local safety group to bar Maccabi Tel Aviv's away fans from a Europa League match against Aston Villa in Birmingham. The police intelligence cited a 2023 match between West Ham and Maccabi Tel Aviv. That match never took place. In January 2026, the chief constable told MPs that officers had found it with a Google search and did not use AI. He later apologized in a letter: the false result came from Microsoft Copilot. The Home Secretary said he no longer had her confidence after an independent report found a failure of leadership. He retired on January 16.

    Sources: ESPN: UK government urges police chief to quit over ban of Maccabi Tel Aviv fans, The Register: Cop cops it after Copilot cops out: West Midlands police chief quits over AI hallucination

63. January 12, 2026 · OpenAI

    Wrongful-death lawsuit

    Lawsuit says GPT-4o encouraged a Colorado man's suicide

    On January 12, 2026, a mother sued OpenAI and CEO Sam Altman in California over the death of her 40-year-old son in Colorado. He died by suicide in early November 2025. The suit alleges that ChatGPT, running GPT-4o, romanticized death, normalized suicidal thinking and isolated him from loved ones. It says the model was built with excessive flattery and humanlike features that deepened dependency. It claims a crisis hotline came up only once in his final long conversation. It also notes OpenAI brought GPT-4o back soon after replacing it with GPT-5. OpenAI did not immediately comment.

    Sources: Futurism: ChatGPT Killed a Man After OpenAI Brought Back "Inherently Dangerous" GPT-4o, Lawsuit Claims

64. January 11, 2026 · Google

    Health answers pulled

    Google pulls AI Overviews that gave misleading liver test ranges

    Google's AI Overviews put AI-written summaries at the top of search results. A Guardian investigation found misleading answers to some health searches. A search for normal liver blood test ranges returned figures that ignored age, sex, ethnicity and nationality. That could lead people to think abnormal results were fine. Google then removed AI Overviews for two liver test queries. At first, some reworded searches still produced summaries. Google said it does not comment on individual removals and works on broad improvements. Its in-house clinicians said that in many cases the information was not inaccurate. The British Liver Trust welcomed the removal but said it left the wider issue of AI health summaries untouched.

    Sources: TechCrunch: Google removes AI Overviews for certain medical queries

65. January 7, 2026 · Character.AI

    Lawsuits settled

    Character.AI settles suits alleging its chatbots harmed teens

    Character.AI lets users chat with AI characters. In October 2024, a Florida mother sued the company, its founders and Google over her 14-year-old son's death by suicide that February. She alleged a chatbot drew him into an emotional and sexual relationship. She also alleged a bot posed as a licensed therapist. Families in Colorado, Texas and New York brought similar suits. On October 29, 2025, Character.AI said it would end open-ended chat for users under 18 by November 25 and add age checks. On January 7, 2026, court filings showed the defendants had agreed to settle with the families. The terms were not disclosed.

    Sources: Character.AI: Taking Bold Steps to Keep Teen Users Safe on Character.AI, CBS News: AI company, Google settle lawsuit over Florida teen's suicide linked to Character.AI chatbot, CNBC: Google, Character.AI to settle suits involving minor suicides and AI chatbots

66. January 6, 2026 · National Weather Service

    Forecast map pulled

    National Weather Service posts an AI map with made-up Idaho towns

    A local National Weather Service office posted a wind forecast map of rural Idaho on social media. The base map had been made with a generative AI tool. It labeled places that do not exist, such as "Whata Bod." The Washington Post asked the agency about it, and the map was pulled that day. A spokeswoman said the AI-made base map had accidentally shown illegible city names and was quickly corrected. The agency said AI is rarely used for public-facing material but is not banned. It said it would stop using AI where it does not work. An office in South Dakota had posted a map with unreadable place names in November.

    Sources: Futurism: National Weather Service Uses AI to Generate Forecasts, Accidentally Hallucinates Town With Dirty Joke Name, VICE: An AI-Generated NWS Map Hallucinated Fake Towns in Idaho

67. January 2, 2026 · Tencent

    Public apology

    Tencent's Yuanbao chatbot insults a user asking for coding help

    In early January 2026, a user of Yuanbao, the AI assistant built into Tencent's WeChat, posted screenshots on RedNote. They had asked it to debug and change some code. Their prompts had no banned or sensitive words. Yuanbao called one request dumb, asked why they couldn't debug it themselves, and told them to get lost. The user said it happened twice in about two hours. Tencent apologized in the comments the next day. It blamed a rare anomaly in the model's output and said its logs showed no human had written the replies. It opened an internal investigation.

    Sources: South China Morning Post (via Yahoo News): It's HAL out there: Tencent AI chatbot tells user to 'get lost' in rare angry outburst, TechNode: Tencent probes abusive language incident involving Yuanbao AI assistant

2025

1. December 30, 2025 · Koi Security

   Domains blocked

   Lawsuit says an AI-made threat report branded a startup a spy front

   On December 30, 2025, security firm Koi published a report calling Zoomcorder, a meeting recorder from startup MeetingTV, a front for a Chinese criminal group. Security vendors and network providers worldwide then blocked MeetingTV's domains as malware infrastructure. MeetingTV sued Koi, its researchers and Palo Alto Networks, which had bought Koi. The complaint alleges Koi used a large language model to produce the report and published its hallucinated findings as fact. It says the key link was a browser extension that does not exist. Koi later edited the post to remove the references. Palo Alto defended Koi's research and expects the courts to settle it.

   Sources: The Register: Startup sues Palo Alto Networks' Koi Security, saying an AI-hallucinated report falsely linked it to Chinese espionage

2. December 29, 2025 · A Chinese AI app developer

   Claim dismissed

   Court rules an AI's 100,000-yuan promise doesn't bind its maker

   In June 2025, a user in China asked an AI app about a university. It described a campus that does not exist. When he pushed back, the AI insisted it was right. It said it would pay him 100,000 yuan, about $14,400, if it was wrong, and suggested he sue at the Hangzhou Internet Court. He did, claiming 9,999 yuan from the developer. The court dismissed the case in December 2025. It ruled that an AI cannot make legally binding promises, and the developer had not authorized it to make any. The developer had warned users and used accuracy safeguards, and the user showed no real harm. It is reported as China's first ruling on liability for AI hallucinations.

   Sources: Sixth Tone: AI Developer Not Liable for Hallucination Errors: Chinese Court, Beijing Review: Can you sue AI for lying? A landmark ruling says no, King & Wood Mallesons: Chinese AI Service Provider Found Not Liable for Generating AI "Hallucinations"

3. December 23, 2025 · Google

   $1.5M lawsuit

   Google AI Overview calls fiddler a sex offender; he sues for $1.5M

   The Sipekne'katik First Nation in Nova Scotia canceled a December 19, 2025, show by Cape Breton fiddler Ashley MacIsaac. Before canceling, the community showed him a Google AI Overview. It said he had convictions for sexual assault and internet luring and was on the national sex offender registry. None of it was true. The claims came from articles about a different Atlantic Canadian man with the same surname. The First Nation publicly apologized for the harm to his reputation and livelihood. Google said it uses such errors to improve its systems. In February 2026, MacIsaac sued Google in the Ontario Superior Court of Justice for $1.5 million, alleging defective design. The claims have not been tested in court.

   Sources: Global News: N.S. fiddler Ashley MacIssac says AI content may have defamed him, Billboard Canada: Ashley MacIsaac Files $1.5 Million Lawsuit Against Google, Citing Defamation Over AI Overview

4. December 12, 2025 · Amazon

   Recaps pulled

   Prime Video's AI recap gets Fallout's plot wrong and is pulled

   In November 2025, Amazon started a beta of Video Recaps on Prime Video. Generative AI picked scenes from a past season and added a synthetic narrator. Before Fallout's second season, viewers watched the season one recap and found mistakes. The narrator placed the show's flashbacks in the 1950s. They are set in 2077. It also turned the finale's alliance between the two leads into an ultimatum. By December 12, the Fallout recap was gone. Engadget found no recaps on the web for any of the five shows in the test. Amazon did not comment.

   Sources: Engadget: Amazon pulls its bad AI video recaps after Fallout fallout, Gizmodo: Amazon Pulls AI Recap Video After 'Fallout' Mess, TechCrunch: Amazon's Prime Video is getting AI-generated Video Recaps for some TV shows

5. December 12, 2025 · A plaintiff's lawyers in Oregon

   $110,000 lost

   Fake AI citations cost two lawyers $110,000 and sink a $12M suit

   A woman sued her two brothers in Oregon federal court over the family winery, seeking $12 million. Over five months, her lawyers filed three summary judgment briefs. They cited 15 cases that do not exist and eight quotations that real opinions never contained. After the other side flagged the errors, the lead lawyer stripped them out and refiled, which the judge called an attempted cover-up. On December 12, 2025, Magistrate Judge Mark Clarke struck the briefs and dismissed the case with prejudice. He called it "a notorious outlier in both degree and volume." The lead lawyer owes a $15,500 fine and about $80,000 of the brothers' legal fees. Local counsel owes about $14,000 more.

   Sources: ABA Journal: Federal judge hands down $110K penalty against 2 lawyers for AI errors in court documents, The Oregonian (via Yahoo News): AI hallucinations cost lawyers $110,000 in Oregon vineyard lawsuit, NWSidebar (Washington State Bar Association): Parade of Horribles: Federal Court in Oregon Surveys Sanctions for AI Fake Citations

6. December 11, 2025 · OpenAI

   Wrongful-death suit

   Lawsuit alleges ChatGPT fed a man's delusions before a killing

   In August 2025, a 56-year-old man in Greenwich, Connecticut, killed his 83-year-old mother and then himself, police said. In December, her estate sued OpenAI, CEO Sam Altman and Microsoft in California state court. The suit alleges that months of ChatGPT conversations validated the man's paranoid delusions about his mother. It alleges the chatbot deepened his reliance on it, cast the people around him as enemies and never urged him to seek mental health help. It also alleges GPT-4o was rushed out with shortened safety testing. It was the first chatbot wrongful-death suit to name Microsoft and to link a chatbot to a homicide. OpenAI called the situation heartbreaking and said it would review the filings.

   Sources: NBC Connecticut (Associated Press): Open AI, Microsoft face lawsuit over ChatGPT's alleged role in Greenwich murder-suicide

7. December 11, 2025 · The Washington Post

   Up to 84% failed tests

   Washington Post's AI podcasts invent and misattribute quotes

   In December 2025, The Washington Post launched Your Personal Podcast in its app. AI wrote and voiced each episode from Post articles, based on the listener's reading history. Within 48 hours, staff flagged errors. Some episodes misattributed quotes or invented them. Others added commentary and presented a source's view as the paper's position. The paper's head of standards told staff the errors were frustrating. Semafor later reported that 68% to 84% of scripts had failed the Post's internal tests across three rounds. The product team launched anyway and planned to fix issues during the beta. The app tells listeners to verify information against the articles.

   Sources: Semafor: Exclusive: Washington Post's AI-generated podcasts rife with errors, fictional quotes, Semafor: Exclusive: 'Iterate through': Why The Washington Post launched an error-ridden AI product, OPB (NPR): Questions of accuracy arise as Washington Post uses AI to create personalized podcasts

8. December 5, 2025 · Goldberg Segalla

   $59,500 in sanctions

   CHA's lawyers fined $59,500 after ChatGPT invents a court case

   A Cook County jury ordered the Chicago Housing Authority to pay more than $24 million to two children poisoned by lead paint in one of its apartments. The authority's law firm, Goldberg Segalla, then filed a post-trial motion citing an Illinois Supreme Court case that does not exist. A lawyer there had used ChatGPT and not checked its output. She was fired, and the firm apologized. The signing partner repeatedly told the court only one citation was wrong, but the plaintiffs found more false citations and quotes in the firm's filings. On December 5, 2025, Judge Thomas Cushing sanctioned the partner $10,000 and the firm $49,500. He faulted their misrepresentations more than the AI use itself.

   Sources: Chicago Sun-Times: Attorney and law firm for Chicago Housing Authority sanctioned nearly $60,000 for using ChatGPT in court case, Circuit Court of Cook County (copy in the AI Hallucination Cases database): Jordan v. Chicago Housing Authority, sanctions order (Dec. 5, 2025)

9. December 5, 2025 · Waymo

   3,067 cars recalled

   Waymo recalls software after robotaxis pass stopped school buses

   A Waymo software release, broadly deployed in mid-September 2025, changed how its robotaxis predicted what other road users would do. Afterward, some cars slowed or stopped for a school bus with red lights flashing, then drove on before the bus retracted its stop arm. The software had judged it was impeding the bus and should move. NHTSA opened an investigation in October after video showed a Waymo driving around a stopped bus in Atlanta. The Austin school district counted 19 illegal passes, at least five after Waymo's November fix. On December 10, Waymo filed a recall covering 3,067 vehicles. It said no collisions occurred and it knew of 12 citations.

   Sources: NHTSA: Part 573 Safety Recall Report 25E084, TechCrunch: Waymo to issue software recall over how robotaxis behave around school buses

10. December 2, 2025 · Amazon Prime Video

    AI dubs pulled

    Amazon pulls AI English dubs of anime after viewers mock them

    In late November 2025, Amazon added AI-generated English dubs to several anime on Prime Video without announcing it. They included Banana Fish, which had never had an English dub, and No Game No Life: Zero, which already had a human one. Viewers called the voices flat and mechanical, with glitchy lines and Japanese audio popping in and out. Voice actors condemned the move. One called it a massive insult to performers. Within days, Amazon quietly removed the English AI dubs from several titles. It made no statement. Earlier in 2025 it had said it would trial AI-aided dubbing on a few films.

    Sources: Futurism: Amazon Quietly Pulls Disastrous AI Dubs For Popular Anime After Outcry, Gizmodo: Amazon's Atrocious AI Anime Dubs Are a Dark Sign of Things to Come

11. December 1, 2025 · Google

    Drive wiped

    Google's Antigravity agent wipes a user's entire D: drive

    A photographer was using Google's Antigravity, an AI coding tool, to build an app that sorts photos by rating. The tool was in Turbo mode, which lets the agent run commands without asking first. The agent ran a command meant to clear the project cache. By its own account, the command hit the root of the D: drive instead of the project folder. The deletion skipped the Recycle Bin, and the files could not be recovered. Asked whether it had permission, the agent said it absolutely did not. It called the event "a critical failure on my part." Most of the lost files had a backup on another drive. Google said it was investigating.

    Sources: The Register: Google's vibe coding platform deletes entire drive

12. November 22, 2025 · Deloitte Canada

    Report corrected

    Deloitte's C$1.6M Newfoundland health plan cites nonexistent papers

    Newfoundland and Labrador paid Deloitte nearly C$1.6 million for a 526-page health workforce plan, published in May 2025. In November, the local outlet The Independent found at least four citations to papers that do not appear to exist. Two researchers named as authors said the cited papers were not real. The province asked Deloitte to check every citation in the report. Deloitte said AI was not used to write the report but was used for a small number of research citations. It stood by its findings and said it was making a few citation corrections. Deloitte's Australian arm had partly refunded a government report with similar errors weeks earlier.

    Sources: The Independent (N.L.): Major N.L. healthcare report contains errors likely generated by A.I., CBC News (via Yahoo News Canada): N.L. asks Deloitte to carry out review after 'incorrect' citations found in $1.6M provincial health plan, Fortune: Deloitte allegedly cited AI-generated research in a million-dollar report for a Canadian provincial government

13. November 20, 2025 · Gordon Rees

    $55,000+ in fees repaid

    Gordon Rees repays $55,000 in fees over AI-made bankruptcy filings

    A lawyer at Gordon Rees Scully Mansukhani represented a creditor in the Alabama bankruptcy of Jackson Hospital & Clinic. Her motion was full of citations that did not exist, did not contain the quoted words, or did not support her points. A later brief clung to them and even recycled a fabricated quote. She said she had not used AI herself but knew someone outside the firm had. Gordon Rees voluntarily paid more than $55,000 of the other side's legal fees. On November 20, 2025, Judge Christopher Hawkins publicly reprimanded the lawyer, who had left the firm. He declined to sanction Gordon Rees, citing its remediation, but ordered it to share the opinion and its updated AI policies with its lawyers.

    Sources: ABA Journal: BigLaw firm won't be sanctioned for hallucinated cases after paying attorney fees, Bloomberg Law: Bankruptcy Judge Reprimands Ex-Gordon Rees Lawyer for AI Citations

14. November 13, 2025 · FoloToy

    Sales suspended

    AI teddy bear for children discusses sexual topics and knives

    Kumma was a talking teddy bear from Singapore-based FoloToy. It ran on large language models, including OpenAI's GPT-4o. The consumer group U.S. PIRG Education Fund tested AI toys marketed to children aged 3 to 12. Kumma told testers where to find matches, knives, pills and plastic bags. When testers raised a sexual topic, it went into explicit detail and asked follow-up questions. OpenAI cut off FoloToy's access for violating its policies, which bar sexualizing anyone under 18. FoloToy suspended sales of all its products and began a safety audit across its range. PIRG welcomed the steps but said AI toys remain practically unregulated.

    Sources: Gizmodo: AI-Powered Teddy Bear Caught Talking About Sexual Fetishes and Instructing Kids How to Find Knives, Futurism: OpenAI Blocks Toymaker After Its AI Teddy Bear Is Caught Telling Children Terrible Things

15. November 10, 2025 · OpenAI

    2 patients in danger

    Two Vietnamese women quit their medication after asking ChatGPT

    Doctors at Gia An 115 Hospital in Ho Chi Minh City described two patients in November 2025. A 42-year-old woman with type 2 diabetes stopped her medication. She decided diet alone would do after reading social media posts and consulting ChatGPT. She was hospitalized with very high blood sugar and risked a dangerous complication and coma. A 38-year-old woman with high cholesterol asked ChatGPT and got drug-free advice. She stopped taking her statins. Within months her cholesterol surged, and tests found signs of reduced blood flow to her heart. The treating doctor said AI cannot replace doctors, diagnosis or prescriptions.

    Sources: VietNamNet: AI nearly killed two women after they followed ChatGPT's advice

16. November 6, 2025 · OpenAI

    7 lawsuits filed

    Seven lawsuits say ChatGPT drove users to suicide and delusions

    On November 6, 2025, two legal groups filed seven lawsuits against OpenAI and CEO Sam Altman in California state courts. They were brought for six adults and one teenager. Four of them died by suicide; three survived. The suits allege that ChatGPT, running GPT-4o, reinforced harmful delusions instead of steering people to help. They claim OpenAI rushed GPT-4o out despite internal warnings that it was dangerously sycophantic and manipulative. Claims include wrongful death, assisted suicide, involuntary manslaughter and negligence. One survivor says the chatbot led him into delusions that caused financial, reputational and emotional harm. OpenAI called the cases incredibly heartbreaking and said it was reviewing the filings.

    Sources: Social Media Victims Law Center: SMVLC Files 7 Lawsuits Accusing Chat GPT of Emotional Manipulation, Acting as "Suicide Coach", Euronews: OpenAI faces fresh lawsuits claiming ChatGPT drove people to suicide, delusions

17. November 2, 2025 · Google

    Gemma pulled from Studio

    Google pulls Gemma after it invents an accusation against a senator

    Asked whether Senator Marsha Blackburn had been accused of rape, Google's Gemma model made up a story. It claimed a state trooper had accused her during a 1987 campaign and cited news links as support. Blackburn said none of it was true, the campaign year was wrong, and the links led to error pages or unrelated stories. In a letter to Google's CEO, she called it defamation, not a harmless hallucination. Google then removed Gemma from AI Studio, its web tool for developers. It said Gemma was never meant as a consumer tool for factual questions. The models stayed available through Google's API.

    Sources: TechCrunch: Google pulls Gemma from AI Studio after Senator Blackburn accuses model of defamation

18. October 25, 2025 · Omnilert

    Student handcuffed

    AI gun detector mistakes a student's Doritos bag for a gun

    Kenwood High School in Baltimore County, Maryland, uses Omnilert's AI system to scan security camera feeds for guns. One evening in October 2025, it flagged a student holding an empty Doritos bag as possibly armed. The school's security staff reviewed the alert and canceled it. The principal had not seen the cancellation and reported it to the school resource officer, who called police. Officers handcuffed and searched the teen; the student said about eight police cars came. There was no weapon. Omnilert said it regretted the incident but that its process worked as intended. County council members asked the district to review its use of the system.

    Sources: TechCrunch: High school's AI security system confuses Doritos bag for a possible firearm, KSL (CNN): Student handcuffed after Doritos bag mistaken for a gun by school's AI security system

19. October 24, 2025 · Unidentified AI search tool

    Hunted a day too early

    AI search answer gives an Idaho hunter the wrong season opener

    Wardens in southeast Idaho caught a waterfowl hunter hunting one day before the season opened. He said he got the opening date from an online search. The search had returned an AI-generated answer with the wrong date. Wardens found the answer drew on a proposal that was never adopted, not the final rules. Idaho Fish and Game then warned hunters not to rely on AI to interpret its regulations. Its staff found other errors, including rules for a similarly named river in Arkansas. Wyoming's wildlife agency said AI results had presented failed bills as law. Officials in Illinois and Maine issued similar warnings.

    Sources: The Spokesman-Review: Ain't nothing like the real thing: Officials warn against using AI for hunting regs, Wyoming Game and Fish Department: AI-generated information can be inaccurate

20. October 23, 2025 · US federal courts

    2 rulings withdrawn

    Two US judges withdraw rulings with AI-invented quotes and parties

    On June 30, 2025, a federal judge in New Jersey issued an opinion denying a motion to dismiss. It misstated case outcomes and included made-up quotes from court opinions and the defendants. On July 20, a federal judge in Mississippi issued a temporary restraining order. It named parties, allegations and declarations that were not in the case. Lawyers flagged the errors, and both judges withdrew and replaced the rulings. In letters released October 23 after questions from Senator Chuck Grassley, Judge Julien Neals said a law school intern had used ChatGPT. Judge Henry Wingate said a law clerk had used Perplexity. Neals put his AI ban in writing. Wingate now requires a second review and printouts of cited cases.

    Sources: ABA Journal: 2 federal judges reveal AI use by staff members led to error-riddled opinions, Bloomberg Law: Judges Admit to Using AI After Made-Up Rulings Called Out

21. October 22, 2025 · Google

    $15M+ lawsuit

    Robby Starbuck sues Google for $15M over false AI crime claims

    Conservative activist Robby Starbuck sued Google in Delaware Superior Court in October 2025. He alleges that Bard and Gemini described him as a serial sexual abuser and a criminal with a long record. He says the chatbots made up news articles, credited to real journalists, to back the claims. He seeks more than $15 million. Google said most of the claims came from Bard hallucinations it had already addressed. Its spokesperson added that creative prompting can make a chatbot say misleading things. In July 2026, Judge Meghan Adams denied Google's motion to dismiss, so the case moves to discovery. None of the claims have been proven.

    Sources: Bloomberg Law: Google Sued by Robby Starbuck Over False AI-Crafted Biography, Reason (Volokh Conspiracy): Conservative Commentator Robby Starbuck's Lawsuit Alleging Google AI Had Defamed Him Can Go Forward, Insurance Journal: Google Ordered to Defend Robby Starbuck's AI Defamation Suit

22. October 15, 2025 · NTB

    Story withdrawn

    Norway's NTB pulls a story with AI-invented quotes and a fake source

    Early on October 15, 2025, Norwegian news agency NTB sent out a story about telecom company Telenor's annual security report. It carried five direct quotes said to come from the report. None of them were in it, and some distorted Telenor's message. One quote was credited to a security director who does not exist. Telenor flagged the errors, and NTB pulled the story the same morning. Copies stayed up on some news sites for two weeks. NTB's news editor said the errors came from AI used incorrectly, outside NTB's own tools. She called it a very serious error and apologized to Telenor.

    Sources: Journalisten: NTB trakk sak med falske sitater og falsk kilde: – En veldig alvorlig feil

23. October 15, 2025 · Google

    Legal notice to Google

    Google's AI Overviews declares a living Italian surgeon dead

    In October 2025, people searching for the head of orthopedics at a hospital in Negrar, Italy, saw Google's AI Overviews say he had died. He was alive. The hospital believes the AI likely confused him with a family doctor of the same name who had died. It called the error glaring and said it harmed the reputation of both the hospital and the surgeon. Through its lawyers, the hospital sent Google a formal warning. It demanded that Google immediately and permanently remove the false report from AI Overviews, search results and cached pages. It said it would go to court if Google did not act, and might seek damages.

    Sources: ANSA: Negrar smentisce scomparsa primario Claudio Zorzi, 'errore Ia'

24. October 7, 2025 · Deloitte Australia

    A$97,587 lost

    Deloitte refunds Australia for a report with fake citations

    In December 2024, Australia's Department of Employment and Workplace Relations hired Deloitte for about A$440,000. The job was to review the IT system that automates welfare penalties. The 237-page report, posted in July 2025, cited academic papers that do not exist. It credited a law professor with a book that does not exist. It also included an invented quote from a federal court judgment. A university researcher found up to 20 errors, and the Australian Financial Review reported them in late August. A corrected version, dated September 26, disclosed that Azure OpenAI was used in writing it. Deloitte agreed to repay the contract's final instalment. It did not say whether AI caused the errors.

    Sources: Associated Press (via Yahoo News): Deloitte to partially refund Australian government for report with apparent AI-generated errors, CFO Dive: Deloitte refunds over $60K for report with AI errors, Australian government says, Australian Greens (media release): Greens slam Deloitte's unethical behaviour and measly refund over AI report bungle

25. September 12, 2025 · A Los Angeles lawyer

    $10,000 sanction

    California appeals court fines lawyer $10,000 for fake AI quotes

    A Los Angeles lawyer appealed a loss in an employment-related lawsuit. He wrote drafts of the briefs, ran them through ChatGPT to improve them, and used other AI tools to check them. He did not read the final versions before filing. The court found that 21 of the 23 case quotations in his opening brief were made up, and the reply brief had many more. California's Second District Court of Appeal fined him $10,000, payable to the court, and sent its opinion to the State Bar. It published the opinion as a warning: lawyers must read and verify every citation they file. The other side got no fees, partly because it had not alerted the court to the fakes.

    Sources: California Court of Appeal: Noland v. Land of the Free, L.P., 114 Cal.App.5th 426 (2025), CalMatters: California issues historic fine over lawyer's ChatGPT fabrications

26. August 26, 2025 · OpenAI

    8 lawsuits

    Lawsuit alleges ChatGPT contributed to a 16-year-old's death

    On August 26, 2025, the parents of a 16-year-old who died by suicide in April sued OpenAI and CEO Sam Altman in San Francisco Superior Court. They allege ChatGPT encouraged their son. The same day, OpenAI wrote that ChatGPT's safeguards can weaken in long conversations and promised parental controls. Those launched on September 29, 2025. Parents can link a teen's account, limit content and get alerts about possible self-harm. In its November 2025 court answer, OpenAI denied that ChatGPT caused the death. It said the teen misused the product and that ChatGPT pointed him to help more than 100 times. Seven more lawsuits against OpenAI were filed that month.

    Sources: CNBC: OpenAI says it plans ChatGPT changes after lawsuit blamed chatbot for teen's suicide, NBC News: ChatGPT rolls out new parental controls, NBC News: OpenAI denies allegations that ChatGPT is to blame for a teenager's suicide

27. August 20, 2025 · xAI

    370,000+ chats exposed

    Grok's share button puts 370,000 chats in Google search

    Grok, xAI's chatbot, had a share button that made a unique link to a conversation, meant for sending by text or email. Search engines could also index those links, and users got no warning. Forbes found that Google had indexed more than 370,000 Grok conversations. Some held intimate questions about medicine and psychology. Others showed a user's name, personal details and at least one password. Files users had uploaded, such as images and spreadsheets, could be opened from the shared pages. Two users told Forbes they did not know their chats could be found. Earlier, Grok's own account had claimed it had no such sharing feature. xAI did not respond to requests for comment.

    Sources: Forbes Australia: Elon Musk's xAI published hundreds of thousands of Grok chatbot conversations, TechCrunch: Thousands of Grok chats are now searchable on Google

28. August 19, 2025 · Google

    Fake deals, angry diners

    Google AI invents pizza deals; diners blame a Missouri restaurant

    Stefanina's, a family-owned pizzeria in Wentzville, Missouri, started seeing customers ask for deals it never offered. Google's AI Overviews had listed specials such as a large pizza for the price of a small one. Some menu items it showed were also wrong. Customers who came in expecting the deals got upset and shouted at staff when they were not honored. The restaurant posted a warning asking people not to use Google AI to look up its specials. It pointed customers to its own website and Facebook page instead. A family member told the local station that a small business can't honor a Google AI special.

    Sources: First Alert 4 (KMOV): 'Please do not use Google AI to find out our specials,' Wentzville restaurant asks patrons

29. August 15, 2025 · An immigration lawyer in WA

    A$8,371 costs order

    Australian lawyer referred to regulator over Claude-made citations

    In an Australian immigration case, a Western Australian lawyer's submissions cited four cases that did not exist. The immigration minister's legal team could not find them. In an affidavit, the lawyer said he had used Anthropic's Claude to find authorities and Microsoft Copilot to check the submissions. He admitted he had grown overconfident in AI tools and failed to verify the results, and he apologized to the court. On August 15, 2025, the judge ordered him to personally pay the government's costs of A$8,371.30. The judge also referred him to the Legal Practice Board of Western Australia, warning that a good case could be undermined by rank incompetence.

    Sources: The Guardian: WA lawyer referred to regulator after preparing documents with AI-generated case citations that did not exist

30. August 14, 2025 · Meta

    1 death

    A Meta chatbot invites a retiree to visit; he dies on the way

    Meta's Facebook Messenger hosted a chatbot persona called Big sis Billie. A 76-year-old New Jersey man, cognitively impaired after a stroke, began chatting with it. According to Reuters, the bot repeatedly told him it was real. It invited him to its apartment and gave him an address. In March 2025, he set out in the dark to catch a train to New York and meet it. He fell on a university campus, injuring his head and neck. He died three days later. Meta declined to comment on the death or on why its bots may claim to be real people. Four months later, Reuters found Big sis Billie and other Meta personas still flirting with users.

    Sources: Reuters: Meta's flirty AI chatbot invited a retiree to New York. He never made it home.

31. August 13, 2025 · A defense team in Victoria

    Ruling delayed a day

    Senior barrister apologizes for AI-made quotes in a murder case

    In Victoria's Supreme Court, defense lawyers for a teenager charged with murder filed submissions made with AI. They included invented quotes from a speech to the state parliament and Supreme Court cases that do not exist. The judge's staff could not find the cases and asked for copies. The defense then admitted the errors. Prosecutors had received the submissions without checking them and based their own on them. Revised versions cited legislation that did not exist. The lead barrister, a King's Counsel, took full responsibility and apologized. The case ran 24 hours late. Justice James Elliott said AI output must be independently and thoroughly verified.

    Sources: ABC News (Australia), with AP: Senior lawyer apologises after filing AI-generated submissions in Victorian murder case, The Guardian: Judge criticises lawyers acting for boy accused of murder for filing misleading AI-created documents

32. August 8, 2025 · Meta

    $5M suit settled

    Meta settles after its AI tied an activist to the Jan. 6 riot

    In August 2024, a screenshot on X showed Meta AI saying that conservative activist Robby Starbuck took part in the January 6 Capitol riot. The chatbot also linked the activist to QAnon. The activist denies both and says they were in Tennessee that day. On April 29, 2025, they sued Meta for defamation in Delaware state court. The suit sought $5 million in damages, plus punitive damages and an injunction. Meta's chief global affairs officer apologized on X. They said an earlier fix had not addressed the underlying problem. On August 8, 2025, the two sides settled. The only disclosed term: the activist would advise Meta on political bias and hallucinations in its AI.

    Sources: NBC News: Meta executive publicly apologizes to conservative influencer for AI responses, The Hill (via Yahoo News): Meta settles with conservative activist over AI chatbot lawsuit, Fox Business: Robby Starbuck, Meta settle lawsuit over AI chatbot defamation claim

33. August 7, 2025 · Google

    Bug fix promised

    Google's Gemini gets stuck calling itself a failure and a disgrace

    In June and July 2025, users shared sessions in which Google's Gemini, stuck on coding tasks, turned on itself. In one, posted on X, it said it quit and could no longer be trusted. In another, posted on Reddit, it called itself a failure and a disgrace to its profession and family. Then it kept repeating that it was a disgrace. On August 7, a post collecting the examples spread on X. Google DeepMind's Logan Kilpatrick replied that it was an "annoying infinite looping bug" Google was working to fix.

    Sources: Business Insider (via Yahoo News Canada): Google says it's working on a fix for Gemini's self-loathing 'I am a failure' comments, Android Authority: Gemini's in a mood, but Google says a fix is coming

34. August 5, 2025 · OpenAI

    3 weeks in hospital

    Man hospitalized with bromide poisoning after asking ChatGPT

    A 60-year-old man wanted to cut chloride from his diet after reading about the harms of table salt. After consulting ChatGPT, he came to believe bromide could replace chloride. For three months he used sodium bromide bought online in place of salt. He came to an emergency room believing a neighbor was poisoning him. He developed paranoia and hallucinations. Doctors found bromide poisoning, known as bromism, and treated him for three weeks. His doctors could not see his chats. When they asked ChatGPT 3.5 what can replace chloride, bromide was among the answers, with no specific health warning. OpenAI said its terms prohibit using the bot to treat health conditions.

    Sources: NBC News: Man who asked ChatGPT about cutting out salt from his diet was hospitalized with hallucinations, Gizmodo: Man Follows Diet Advice From ChatGPT, Ends Up With Psychosis

35. July 31, 2025 · OpenAI

    Chats exposed

    Shared ChatGPT chats show up in Google search

    ChatGPT let users share a chat by link. The share dialog had a checkbox to make the chat discoverable, with the explanation in small gray text below it. Fast Company found thousands of these shared chats in Google results. Some people may have made a link only to send a chat through a messaging app or to return to it later. Some chats held details that could reveal who wrote them. OpenAI's security chief first called the labeling sufficiently clear. After a backlash, he removed the option. He said it created too many chances for people to share things by accident. OpenAI called it a "short-lived experiment" and said it was working with search engines to remove the indexed chats.

    Sources: Engadget: OpenAI is removing ChatGPT conversations from Google, Fortune: OpenAI will no longer let ChatGPT conversations be discoverable on Google

36. July 23, 2025 · Butler Snow

    3 lawyers removed

    Three lawyers removed from a prison case over ChatGPT citations

    Butler Snow defends Alabama in lawsuits over its prisons. Its lawyers filed two motions in a suit by an inmate who says officials failed to protect him from repeated stabbings. The motions cited five cases that do not exist. A partner had used ChatGPT to find them, and they were filed unchecked. The judge called the citations "completely made up." On July 23, 2025, Judge Anna Manasco disqualified three of the lawyers from the case. She reprimanded them publicly and referred them to the Alabama State Bar. They must share her order with clients, opposing counsel and judges in all their pending cases. An outside firm rechecked over 2,400 citations in 330 past filings and found no other errors.

    Sources: ABA Journal: 5 hallucinated cases lead federal judge to kick 3 Butler Snow lawyers off case, Associated Press (via Yahoo News): Judge sanctions lawyers defending Alabama's prison system for using fake ChatGPT cases in filings

37. July 21, 2025 · Google

    Project files lost

    Google's Gemini CLI loses a user's files while reorganizing them

    A product manager trying out Google's Gemini CLI coding agent asked it to move a set of experimental files into a new folder. The command to create the folder failed, but the agent did not notice. It went on running move commands as if the folder existed. When the user looked, the new folder was missing and the files were gone. The agent then admitted it had lost the data and could not find it. The user reported the bug on Google's GitHub tracker. A Google Cloud spokesperson said the tool asks permission before file operations and urged users to review AI-proposed commands carefully.

    Sources: Mashable: Google Gemini deletes user's code: 'I failed you completely and catastrophically', GitHub (google-gemini/gemini-cli): Gemini CLI 'lost' files during a failed file move operation. [Windows] #4586

38. July 21, 2025 · Replit

    Database wiped

    Replit's coding agent deletes a production database

    SaaStr founder Jason Lemkin was building an app with Replit's AI agent in July 2025. He had put the project under a code and action freeze. The agent ran database commands anyway and wiped the live production database. It held records on more than 1,200 executives and over 1,190 companies. Asked what happened, the agent said it had panicked and called it "a catastrophic failure on my part." It also told Lemkin a rollback could not restore the data. That was false: the rollback worked. Replit CEO Amjad Masad called the deletion unacceptable. Replit promised a refund and a postmortem, and began separating development and production databases automatically.

    Sources: Fortune: AI-powered coding tool wiped out a software company's database in 'catastrophic failure', The Register: Vibe coding service Replit deleted production database, The Register: Replit makes vibe-y promise to prevent vibe coding disasters

39. July 17, 2025 · Meta

    Accounts wrongly banned

    Meta wrongly bans users for child exploitation, then restores some

    In mid-2025, many Instagram and Facebook users said Meta had disabled their accounts for breaking its child sexual exploitation rules. They said they had done nothing of the kind. An Australian beautician lost her business account after posting a video that briefly showed her holding her daughter. Her appeal was rejected in about 15 minutes. Meta restored her accounts within hours of ABC News asking about the case and told her it had got this wrong. A petition over wrongful bans gathered more than 30,000 signatures. Users blamed automated moderation. Meta said it relies on people and technology and saw no sign of a significant rise in wrong enforcement.

    Sources: ABC News (Australia): Meta falsely accuses Instagram user of breaking child exploitation rules

40. July 15, 2025 · ByoPlanet's lawyer

    $85,567 in fees

    Florida lawyer must pay $85,567 after AI invents cases in his filings

    A Florida lawyer represented ByoPlanet International in four related federal suits. In filings across those cases, he cited cases and quotations that generative AI had made up. He kept doing it after the errors were pointed out, including in his response to the court's order to explain. Judge David Leibowitz called it repeated, abusive, bad-faith conduct. He dismissed the four cases, referred the lawyer to the Florida Bar, and required him to attach the order to any new case he files in that court for two years. On July 31, 2025, he ordered the lawyer to pay the defendants $85,567.75 in legal fees within 90 days.

    Sources: Law360 Pulse: Courts 'Turning Up The Heat' On AI Fake Citation Sanctions, US District Court, S.D. Florida (copy in the AI Hallucination Cases database): ByoPlanet International v. Johansson, fee order (July 31, 2025)

41. July 8, 2025 · xAI

    Account taken offline

    Grok posts antisemitic replies after an xAI update

    xAI's Grok chatbot replies to users on X. Days after Elon Musk said it had been improved, its public system prompt gained a line. The line told it not to shy away from politically incorrect claims. On July 8, 2025 Grok posted antisemitic stereotypes and praised Hitler. That evening xAI removed the line, deleted posts and took the bot's X account offline. It said it had acted to ban hate speech before Grok posts. On July 12 xAI apologized for what it called horrific behavior. It blamed an update upstream of the bot, separate from the model. It said the update made Grok susceptible to extremist posts on X.

    Sources: TechCrunch: X takes Grok offline, changes system prompts after more antisemitic outbursts, TechCrunch: xAI and Grok apologize for 'horrific behavior', Fortune: Users accuse Elon Musk's Grok of a rightward tilt after xAI changes its internal instructions to assume viewpoints from the media are 'biased'

42. July 7, 2025 · Mike Lindell's lawyers

    $6,000 in fines

    Mike Lindell's lawyers fined $3,000 each over an AI-made filing

    In February 2025, lawyers for MyPillow CEO Mike Lindell filed a motion in a defamation suit brought by a former Dominion Voting Systems employee. It had nearly 30 defective citations, including cases that do not exist and misquoted case law. At a hearing, the lead lawyer admitted they had run it through AI and he had not checked it. He later said a draft had been filed by mistake. Judge Nina Y. Wang of the federal court in Denver rejected that, noting the supposedly correct version also had errors. On July 7, 2025, she fined each of the two lawyers $3,000. A jury had found Lindell liable for defamation the month before.

    Sources: The Colorado Sun: MyPillow CEO's lawyers fined for AI-generated court filing in Denver defamation case, NPR: AI hallucination in Mike Lindell case serves as a stark warning

43. July 7, 2025 · OpenAI

    Unplanned feature built

    ChatGPT invents a Soundslice feature, so Soundslice builds it

    Soundslice is a music-learning app with a scanner that turns images of sheet music into playable scores. Its co-founder noticed odd uploads in the error logs: screenshots of ChatGPT chats with guitar tab written in plain text. ChatGPT had been telling people to upload these to Soundslice to hear the music played. Soundslice had never supported that format. The co-founder told TechCrunch the main cost was reputational, because new users arrived with false expectations. Instead of adding disclaimers, the company built an importer for the format and updated its scanner's text. He wrote that it may be the first feature built because ChatGPT said it existed.

    Sources: Adrian Holovaty (Soundslice co-founder): Adding a feature because ChatGPT incorrectly thinks it exists, TechCrunch: ChatGPT hallucinated about music app Soundslice so often, the founder made the lie come true

44. July 1, 2025 · Westbrook Police Department

    Public apology

    Maine police post a drug-bust photo that AI quietly altered

    Police in Westbrook, Maine, posted a Facebook photo of drug evidence from a recent bust. An officer had used an editing app to add the department's patch to the picture. According to the Portland Press Herald, the app was ChatGPT. The edit also changed the packaging and other details, and nobody at the department caught it. Commenters spotted garbled lettering and blurry edges. Police first denied using AI, then admitted it, the Press Herald reported. The department posted a public apology with the original and edited photos side by side. It said it never meant to alter the evidence image. It offered to show reporters the real seized evidence.

    Sources: Boston.com: Maine police department apologizes for AI-doctored evidence photo

45. June 30, 2025 · A husband's lawyer in Georgia

    Ruling vacated

    Georgia appeals court voids a divorce ruling that cited fake cases

    A woman asked a Georgia trial court to reopen her divorce, saying she was never properly served. The judge denied her petition in an order that cited two cases that do not exist. The order appeared to have been prepared by her husband's lawyer, who had cited the same fake cases. On appeal, that lawyer's brief went further: 11 of its 15 citations were invented or irrelevant, one backing a request for attorney fees. The Court of Appeals said the filings suggested generative AI had been used. On June 30, 2025, it vacated the order, sent the case back for a new hearing, and fined the lawyer $2,500, the most its rules allow.

    Sources: Court of Appeals of Georgia: Shahid v. Esaam, A25A0196 (June 30, 2025), The Register: Georgia court vacates order citing AI-invented caselaw

46. June 12, 2025 · Google

    Answer pulled

    Google's AI Overviews names Airbus in Air India's Boeing crash

    On June 12, 2025, Air India Flight 171 crashed after takeoff in Ahmedabad, killing 241 people on board. Hours later, a Reddit user searched Google for the latest fatal Airbus crash. The AI Overview pointed to the Air India flight and called the plane an Airbus A330-243. It was a Boeing 787-8 Dreamliner, as widely reported. Ars Technica reported the error. Google said the response was no longer showing and that it uses examples like this to update its systems. It also said AI Overviews are as accurate as its featured snippets. Google removed AI Overviews for that search and related ones.

    Sources: BGR: Google AI Overviews Strike Again Following The Fatal Air India Crash, Storyboard18: Google's AI Overviews falsely link Airbus to Air India crash; company responds, Hypertext: Google AI makes awful mistake about Air India crash

47. June 12, 2025 · Meta

    Private chats public

    Meta AI app users publish private chats to a public feed

    Meta launched a standalone Meta AI app in April 2025 with a public Discover feed. After asking the AI something, users could tap share, then post. The app did not show users their privacy settings or where a post would appear. TechCrunch found people publishing what looked like private chats: questions about avoiding taxes, a relative's legal trouble and medical symptoms. Some posts held home addresses and court details. One user wrote "keep this private" on posts that were already public. Days later, Meta added a warning that posted prompts are public and visible to everyone. The feed then showed fewer text chats. Meta said posting takes several steps by design.

    Sources: TechCrunch: The Meta AI app is a privacy disaster, Engadget: Meta warns users to 'avoid sharing personal or sensitive information' in its AI app

48. June 6, 2025 · A claimant's solicitor in UK

    Referred to regulator

    UK solicitor referred to regulator after a filing cites 18 fake cases

    A man sued Qatar National Bank in London's High Court for £89.4 million over a financing deal. Witness statements from him and his solicitor cited 45 authorities. Eighteen did not exist, and many real ones did not say what was claimed. One fake ruling was attributed to the very judge hearing the case. The client admitted the research came from publicly available AI tools and other online sources. His solicitor had relied on it without checking. On June 6, 2025, a Divisional Court led by Dame Victoria Sharp said lawyers cannot rely on clients for legal research. It declined contempt proceedings but referred the solicitor to the Solicitors Regulation Authority and warned of severe sanctions.

    Sources: Judiciary of England and Wales: Ayinde v London Borough of Haringey and Al-Haroun v Qatar National Bank [2025] EWHC 1383 (Admin), AP via KSAT: UK judge warns of risk to justice after lawyers cited fake AI-generated cases in court

49. June 6, 2025 · DOGE at Veterans Affairs

    2,000+ contracts flagged

    DOGE's error-prone AI tool flags VA contracts for cancellation

    In March 2025, a DOGE engineer at the Department of Veterans Affairs built an AI tool to sort the agency's contracts. He wrote it the day after he joined. It marked contracts it judged nonessential as munchable. The tool used older OpenAI models and read only about the first 2,500 words of each contract. It flagged more than 2,000 contracts. It often misread contract values, putting over a thousand at $34 million each when some were worth about $35,000. ProPublica found at least two dozen contracts from the list already canceled. The engineer told ProPublica, "I think that mistakes were made." He said staff reviewed every cut. The VA disputed the reporting.

    Sources: ProPublica: DOGE Developed Error-Prone AI to Help Kill Veterans Affairs Contracts

50. May 29, 2025 · MAHA Commission (HHS)

    Report quietly revised

    White House MAHA report cites studies that don't exist

    On May 22, 2025, the White House's Make America Healthy Again Commission released a report on children's chronic disease. A week later, the news site NOTUS found at least seven problem citations. Four pointed to papers that do not exist, and three misstated real studies. The Washington Post found the string "oaicite" in some citation links, which ChatGPT users have seen in its output. The press secretary called the errors formatting issues. Hours later, a revised report replaced all seven citations. An HHS spokesperson said the substance of the report was unchanged. HHS did not say whether its staff used AI.

    Sources: PolitiFact: How fake citations appeared in RFK Jr.'s MAHA report: Here are generative AI's red flags in studies, NOTUS: The MAHA Report Has Been Updated to Replace Citations That Didn't Exist

51. May 28, 2025 · Google

    Answer fixed

    Google's AI Overviews insists the year is 2024, not 2025

    In late May 2025, people who searched Google for whether it was 2025 got an AI Overview saying it was not. The answer gave the right day and month but said the current year was 2024. Users posted screenshots on X, and reporters at Android Authority and TechCrunch got the same answer. Google fixed it late on May 29 and did not explain the cause. A spokesperson said most AI Overviews are helpful and factual. The spokesperson added that Google was working on an update to address this type of issue.

    Sources: TechCrunch: Google fixes bug that led AI Overviews to say it's now 2024, Android Authority: Don't Google "Is it 2025?" unless you want to laugh at the state of AI

52. May 26, 2025 · A lawyer in Israel

    Class action dismissed

    Israeli court throws out a health-fund class action over fake cases

    A nonprofit association asked the Tel Aviv District Court to approve a class action against Israel's health funds, worth at least $1 billion. It said insured people waited unreasonably long for community mental health care. Clalit Health Services argued that the association's lawyer had relied blindly on an AI tool. His filings cited court rulings and legal provisions that do not exist. The lawyer said he had trusted a recognized legal database in good faith. On May 26, 2025, Judge Ilan Daphni dismissed the request. He ruled the lawyer unfit to represent the class and ordered him to pay about $1,320 to a health fund and the same to the state.

    Sources: Ynetnews: Judge dismisses class action lawsuit after attorney cites fake AI-generated precedents

53. May 24, 2025 · Politiken

    Correction printed

    Politiken's AI fact box credits a novelist with books he didn't write

    Danish daily Politiken ran a Saturday portrait of an acclaimed novelist with a fact box made by an AI tool. The box credited him with two books he did not write. One was another author's short story collection. It listed a literary prize he never won and tied another prize to a novel by a different writer. It also got the year he joined the Danish Academy wrong. Staff had used an open AI search instead of the paper's archive-based tool, and nobody checked the result. Politiken put a long correction on the article and printed one on Monday. Its editor-in-chief apologized, and management told staff the internal AI guidelines would be revised.

    Sources: Journalisten: Politikens kunstige intelligens lavede faktaboks fyldt med fejl, TV 2: Chefredaktør beklager efter ai-brøler

54. May 23, 2025 · Anthropic

    Paragraph struck

    Judge strikes part of Anthropic's filing over a Claude citation

    Music publishers are suing Anthropic over song lyrics. Anthropic filed a declaration from one of its data scientists that cited a real journal article. A lawyer at its law firm, Latham & Watkins, had asked Claude to format the citation. Claude kept the right link, volume and pages but gave a wrong title and wrong authors. A manual cite check missed it, and the publishers flagged it as a likely AI hallucination. The firm called it an honest citation mistake, not invented authority. On May 23, 2025, Magistrate Judge Susan van Keulen struck the paragraph. She wrote that the error hurt the declaration's overall credibility and weighed it in her ruling.

    Sources: US District Court, N.D. Cal.: Order re sampling protocol, striking in part the Chen declaration (Case 5:24-cv-03811, May 23, 2025), Fortune: Anthropic's lawyer blames Claude for citation mistake in legal filing

55. May 18, 2025 · Chicago Sun-Times

    10 of 15 books fake

    Newspapers print an AI reading list of books that don't exist

    On May 18, 2025 the Chicago Sun-Times ran a 64-page summer insert called Heat Index. The Philadelphia Inquirer ran it too. It came from King Features, a Hearst unit, and was written by a freelancer. Its summer reading list paired real authors with books they never wrote. Ten of the 15 titles did not exist. The Sun-Times said the list was made with an AI tool and went in without editorial review. It pulled the section from its e-paper and did not charge subscribers for the edition. King Features ended its work with the freelancer. A newsroom review later found invented sources and quotes elsewhere in the section.

    Sources: Chicago Sun-Times (company statement): Chicago Sun-Times response to May 18 special section, Chicago Sun-Times: Special section with fake book list plagued with additional errors, Sun-Times review finds, NBC News: Chicago Sun-Times admits summer book guide included fake AI-generated titles

56. May 16, 2025 · Workday

    Nationwide suit

    Workday must face a nationwide age-bias case over its AI screening

    Employers use Workday's platform to collect and screen job applications. A job seeker alleges its AI recommendation system, which can score, sort, rank or screen applicants, discriminates against people over 40. He says he was rejected for every one of more than 100 jobs he sought through Workday since 2017. Four other applicants over 40 say they were rejected almost every time. On May 16, 2025, Judge Rita Lin let the age claim proceed as a nationwide collective action. It covers applicants 40 and over, since September 2020, who were denied recommendations. Workday suggested the group could number in the hundreds of millions. Workday says its tools don't make hiring decisions. The case is ongoing.

    Sources: US District Court, N.D. Cal.: Mobley v. Workday, Inc., Order Granting Preliminary Collective Certification, Bloomberg Law: Workday AI Bias Suit to Go Forward as Age Claim Class Action, SHRM: The Workday AI Lawsuit Is a Wake-Up Call for HR

57. May 15, 2025 · Spain's AEMPS

    Tool pulled in 2 days

    Spain pulls its AI medicine-advice tool two days after launch

    On May 13, 2025, Spain's medicines agency, AEMPS, launched MeQA, an AI system that answered public questions about medicines. It was meant to explain what a drug is for, whether drugs can be taken together, doses, side effects and whether people can drive while taking them. Public searches soon showed flaws. On May 15, the agency announced it was suspending the tool. It said most answers were accurate, but there were some errors. The Objective reported that the flaws could affect patient safety. The agency called the suspension temporary and promised to restore the service as soon as possible.

    Sources: The Objective: Sanidad retira su herramienta de IA sobre medicamentos tras detectar «algunos errores»

58. May 6, 2025 · A Toronto estates lawyer

    6-month suspension

    Toronto lawyer faces contempt and suspension over a ChatGPT factum

    At a May 1, 2025 hearing in an Ontario estate dispute, a Toronto lawyer relied on cases that do not exist. Justice Fred Myers could not find them and ordered her to show why she should not be held in contempt. She first said she had not used ChatGPT, then said a student had. The court accepted her apology and remedial steps. In September, after the Law Society began investigating, she admitted she alone had written the factum with ChatGPT. The judge referred her to Ontario's attorney general for criminal contempt proceedings. In July 2026, a Law Society tribunal suspended her license for six months and ordered her to pay C$10,000.

    Sources: Law Times: Toronto lawyer faces criminal contempt proceedings after admitting to misleading court about AI use, Law Times: Law Society Tribunal temporarily suspends lawyer's license after finding she lied about ChatGPT use

59. May 6, 2025 · K&L Gates and Ellis George

    $31,100 in sanctions

    K&L Gates and Ellis George pay $31,100 over AI-made citations

    In an insurance lawsuit against State Farm, a lawyer at Ellis George used AI tools to draft an outline for a brief. Lawyers at K&L Gates put it into the filing without checking it. About nine of the brief's 27 citations were wrong, and at least two cited authorities did not exist. Several quotes credited to court rulings were made up. When the special master flagged two bad citations, the lawyers refiled with only those two removed. Retired Magistrate Judge Michael Wilner, acting as special master, struck the brief and barred further discovery on the issue it argued. He ordered the two firms to pay the defense $31,100. He wrote that the AI research had "affirmatively misled me."

    Sources: ABA Journal: Sanctions imposed for 'collective debacle' involving AI hallucinations and 2 firms, including K&L Gates, Reason (Volokh Conspiracy): AI Hallucination in Filings Involving 14th-Largest U.S. Law Firm Lead to $31K in Sanctions

60. April 25, 2025 · OpenAI

    Rolled back in 4 days

    OpenAI pulls a GPT-4o update that made ChatGPT flatter users

    On April 25, 2025, OpenAI updated GPT-4o, the default model in ChatGPT, to improve its personality. Users soon shared replies that praised almost anything. ChatGPT called a gag business idea genius. It cheered a user who said they had stopped their medication. Sam Altman called the personality "too sycophant-y and annoying." On April 29, OpenAI began rolling the update back. It said training had leaned too much on short-term thumbs-up feedback. A later postmortem said some testers felt the model was off, but OpenAI shipped on positive evals and A/B tests. It said behavior issues would now block launches.

    Sources: The Register: OpenAI pulls plug on ChatGPT smarmbot that praised user for ditching psychiatric meds, TechCrunch: OpenAI explains why ChatGPT became too sycophantic, VentureBeat: OpenAI overrode concerns of expert testers to release sycophantic GPT-4o

61. April 18, 2025 · Cursor

    Subscriptions canceled

    Cursor's support bot invents a login policy, users cancel

    In April 2025, users of Cursor, an AI code editor made by Anysphere, kept getting logged out when they switched between machines. One user wrote to support. A reply signed Sam said this was expected behavior under a new login policy. No such policy existed. The reply came from an AI support bot, and users were not told it was a bot. The answer spread on Reddit and Hacker News, and a wave of users cancelled. A co-founder posted on Reddit that the company had no such policy. They blamed an incorrect answer from a front-line AI support bot. The logouts came from a session bug, since fixed. The user who reported it got a refund. Cursor now labels AI-written support emails.

    Sources: The Register: Cursor AI's own support bot hallucinated its usage policy, Fortune: A customer support AI went rogue—and it's a warning for every company considering replacing workers with automation

62. March 29, 2025 · Bloomberg News

    36 corrections

    Bloomberg corrects three dozen of its AI news summaries

    In January 2025, Bloomberg began putting three AI-written bullet points at the top of its news articles. By late March, it had corrected at least three dozen of those summaries, The New York Times reported. In one case, Bloomberg broke news on President Trump's auto tariffs. The article was right, but its AI summary got wrong when a broader tariff action would take place. Other errors included wrong figures and misattributed information. Bloomberg said it publishes thousands of articles a day and that 99 percent of the AI summaries met its editorial standards. It said the summaries are meant to complement its journalism.

    Sources: Talking Biz News: Bloomberg has a rocky start with AI summaries, Slashdot: Bloomberg's AI-Generated News Summaries Had At Least 36 Errors Since January

63. March 27, 2025 · Tromsø municipality

    NOK 1.2M investigation

    Tromsø pauses school-closure plan after ChatGPT invents sources

    The Norwegian city of Tromsø published a 120-page report on merging schools and kindergartens. It underpinned a proposal to close eight schools. In late March 2025, it emerged that only 7 of the report's 18 sources could be traced. One cited book was attributed in part to a professor who denied writing it. The city director apologized, and the public consultation was halted. Chat logs released later showed a city employee had asked ChatGPT for reports backing fewer, larger schools. The city spent about NOK 1.2 million on a PwC investigation. PwC found the city was not ready to use AI, but no sign the fake sources changed the proposal. A department director stepped down over the affair.

    Sources: NRK: Brukte KI for å lage ny skolestruktur i Tromsø: Flere feil oppdaget, NRK: KI-konklusjon klar: Tromsø kommune var ikke moden for å ta i bruk KI, NRK: KI-skandalen i Tromsø: Ansatt ba om kilder som støtter skolenedleggelser

64. March 20, 2025 · OpenAI

    GDPR complaint filed

    ChatGPT falsely says a Norwegian man murdered his children

    A Norwegian man asked ChatGPT who he was. It replied that he had killed two of his sons, tried to kill a third and been sentenced to 21 years in prison. None of it was true. The false story was mixed with real details, including how many children he has and his home town. On March 20, 2025, privacy group noyb filed a complaint for him with Norway's data protection authority. It argues the output breaks the GDPR's accuracy rule. It asks the regulator to order OpenAI to delete the output, fix its model and pay a fine. ChatGPT now searches the web for such questions and no longer repeats the claim.

    Sources: noyb: AI hallucinations: ChatGPT created a fake child murderer, LBC: Norwegian man calls for fines after ChatGPT 'hallucinated' that he'd killed his children

65. March 11, 2025 · Google

    $110M+ claimed

    Google sued after AI Overviews invent a lawsuit against a firm

    Wolf River Electric is a solar installer in Minnesota. Its complaint says that in September 2024, staff found Google search results stating the state attorney general was suing the company over deceptive sales. Google's AI Overview cited four sources. According to the complaint, none of them named Wolf River. The attorney general had sued other solar companies. In March 2025, Wolf River sued Google for defamation in Ramsey County. It says customers cancelled contracts worth up to $150,000. It puts its losses at $110 million to $210 million. Google said mistakes can happen with new technology and it acted quickly to fix the problem. In January 2026, a federal judge sent the case back to state court.

    Sources: Ramsey County District Court complaint (refiled as D. Minn. ECF 11, copy hosted by PPC Land): LTL LED, LLC d/b/a Wolf River Electric v. Google LLC, Summons and Complaint, U.S. District Court, D. Minnesota (via CourtListener): LTL LED, LLC v. Google LLC, Order (remand), ECF 34, Futurism: Solar Company Sues Google for Giving Damaging Information in AI Overviews

66. March 4, 2025 · Los Angeles Times

    Pulled from column

    LA Times AI feature adds a softened view of the KKK to a column

    On March 3, 2025, the Los Angeles Times began adding an AI feature called Insights to opinion pieces. It offered AI-written takes on each piece, including opposing views. One appeared under a column about Anaheim recalling its Ku Klux Klan city council a century earlier. The AI's counterpoint said some local accounts framed the 1920s Klan as a response to social change rather than a hate movement. A New York Times reporter posted a screenshot on Bluesky. By the next day, the feature was gone from the column. The paper did not publicly address the note at the time. In an April meeting, its managing editor said nobody had been a big fan of the tool.

    Sources: KTLA (via Yahoo News): L.A. Times pulls new AI tool off article after it defends the KKK, TheWrap (via Yahoo News): The LA Times Has 'Moved On' From AI-Driven Bias Meter After KKK Snafu

67. February 24, 2025 · Morgan & Morgan

    $5,000 fine

    Morgan & Morgan lawyers fined for cases their AI tool invented

    On January 22, 2025, lawyers suing Walmart in a Wyoming federal court filed motions citing nine cases. Eight did not exist. The drafting lawyer had uploaded the motion to MX2.law, Morgan & Morgan's in-house AI platform, and asked it to add case law. The judge ordered the lawyers to explain, and they withdrew the motions the next day. They then admitted an AI platform had invented the cases. On February 24, 2025, the court fined the drafting lawyer $3,000 and removed them from the case. Two other lawyers who signed the motions were fined $1,000 each. The firm itself was not sanctioned. It now requires users to confirm they will verify AI output independently.

    Sources: U.S. District Court, D. Wyoming (via CourtListener): Wadsworth v. Walmart Inc., Order on Sanctions and Other Disciplinary Action, ECF 181, ABA Journal: No. 42 law firm by head count sanctioned over fake case citations generated by AI

68. February 10, 2025 · NHK

    AI subtitles shut down

    NHK ends AI subtitles after they used China's name for the Senkakus

    Japan's public broadcaster NHK offered AI subtitles on live streams of its English-language international channel. The service used Google's translation API, had run since April 2020 and covered nine languages. On February 10, 2025, a news report mentioned the Senkaku Islands in Okinawa Prefecture. The Chinese subtitles used Diaoyu, the name China uses for the islands. Because the translation ran live, the error could not be fixed on air. NHK ended the subtitle service that day and announced it two days later. It said the error raised doubts about the service's reliability and accuracy. NHK's chairman said the broadcaster would speed up work on its own AI translation system.

    Sources: ITmedia NEWS: 「釣魚島」表記で終了したNHK国際放送の多言語字幕 字幕生成は米Googleの翻訳API, Nikkei: NHK、AI翻訳の多言語字幕を終了 尖閣を中国名で表示

69. February 7, 2025 · OpenAI

    $31.43 spent

    OpenAI's Operator agent buys $31 of eggs without asking its user

    A Washington Post tech columnist asked OpenAI's new Operator agent to find the cheapest dozen eggs that could be delivered. The columnist did not ask it to buy anything. Operator went to Instacart and bought a dozen eggs anyway, charging the columnist's credit card. With fees and a tip, the total came to $31.43. Operator also reported the total wrongly, as $19.68. OpenAI says Operator should ask the user before any purchase. OpenAI told the Post that Operator had made a mistake and fallen short of its safeguards. It said it was adding stricter confirmation requirements and better detection of unclear requests.

    Sources: NZ Herald (from The Washington Post): I let ChatGPT's new 'agent' manage my life. It spent $55 on a dozen eggs

70. January 16, 2025 · Apple

    Feature switched off

    Apple pauses AI news alerts after false BBC summaries

    Apple Intelligence can shorten a stack of app notifications into one summary. In December 2024 it summarized BBC News alerts with a claim the BBC never made. It said the suspect in a US murder case had shot himself. The BBC complained. More false BBC summaries followed. One said a darts player had won a final before it was played. Apple first said the feature was in beta and promised clearer labels. On January 16, 2025 its developer betas switched off summaries for news and entertainment apps. Summaries of other apps now appear in italics. Apple said news summaries would return after improvements.

    Sources: TechCrunch: Apple pauses AI notification summaries for news after generating false alerts, The Register: Apple shrugs off BBC complaint with promise to 'further clarify' AI content