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.