Baidu AI generates false criminal allegations against lawyer, faces defamation lawsuit
Baidu's AI system generated false criminal allegations against lawyer Huang Guigeng by misidentifying and falsely attributing information from unrelated cases. When the false information spread to his clients and their families, Huang sued Baidu for defamation in November 2025, seeking over 1 million yuan in damages and a public apology.
What happened
In May 2025, lawyer Huang Guigeng discovered that Baidu AI had automatically generated several serious false negative statements about him, including allegations of threatening judges, offering bribes, forging seals, and illegally absorbing public deposits.
The false information was generated by Baidu AI and Wenxin Assistant through misidentification, primarily by replacing names in news reports, official documents, and publicly disclosed information, such as changing 'lawyer Huang' or 'party Huang' from unrelated cases directly to 'Huang Guigeng' or falsely attributing information to him.
When the false information was spread to clients and their families, Huang Guigeng brought a defamation lawsuit on November 2, 2025.
Reported impact
Huang Guigeng reported that the false information caused him significant trouble and losses, including potential harm to his professional reputation and client relationships.
He sought damages of 1 million yuan for loss of professional reputation and 50,000 yuan for emotional distress, and demanded that Baidu cease the infringing conduct and publish a clarification and apology.
Index analysis
The false information primarily originated from the AI model's misidentification and recombination of information from unrelated cases, directly substituting names or falsely attributing facts to Huang Guigeng without verification.
Baidu argued that AI hallucination is an unavoidable stage problem in product development and that AI is neither a natural person nor a legal entity, claiming it provides neutral technical services and bears no subjective fault.
Governance controls
Huang Guigeng argued that 'AI hallucination' cannot serve as a defense and that 'technical neutrality' does not absolve platforms of review responsibilities.
He noted that other AI platforms including Douyin, Tencent Yuanbao, Alibaba Qianwen, and DeepSeek did not generate similarly serious false content when tested with the same queries.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.