Burhan Doğuş Ayparlar

← All decisions

China · 12 March 2026

Final

Baidu held liable in defamation for an AI answer saying a lawyer had been jailed

The courts ordered Baidu to apologise in writing for an 'AI smart answer' that showed a lawyer as a convict; its hallucination defence failed.

Court / authority
Nanjing Intermediate People's Court
Date
12 March 2026
Case / decision no.
Li v. Beijing Baidu Netcom Science and Technology Co., Ltd. (李某诉北京百度网讯科技有限公司)
Status
Final
Nanjing Intermediate People's Court, Li v. Beijing Baidu Netcom Science and Technology Co., Ltd. (李某诉北京百度网讯科技有限公司), 12 March 2026 — https://www.burhandogusayparlar.com/ai-decisions/nanjing-baidu-ai-answer-defamation/
  1. Facts
  2. Question
  3. Decision
  4. Why it matters

Facts

The judgment is unpublished; this account rests on press reports. In September 2024 a lawyer practising in Nanjing noticed that Baidu search was showing content about him being convicted. On 30 September 2024 a search asking how many years he had been sentenced to produced a Baidu 'AI smart answer' stating he had been jailed for three years for causing an explosion, illustrated with a photo of him in his lawyer's robe. He sued Baidu for defamation in the Nanjing Jiangbei New Area People's Court, also challenging autocomplete suggestions and 'related searches'. Baidu argued that AI hallucinations are unforeseeable and that the lawyer had used leading queries.

Question

The courts had to decide whether a search engine operator can be liable for a false, derogatory answer generated by its AI that clearly targets a real person. Baidu argued that with current technology hallucinations cannot be fully prevented, so it could not be at fault. The courts also considered whether functions that merely list other pages, such as autocomplete and 'related searches', differ in liability from an AI answer that itself combines text and images. If liability was found, the appropriate remedy was also in dispute.

Decision

The first-instance court found no infringement in the autocomplete and related-search suggestions, where Baidu merely provided search without processing the data. The AI answer, however, was defamatory: it clearly targeted the claimant, used derogatory terms such as 'sentenced', 'defendant' and 'three years' imprisonment', and combined with his photo it lowered his standing. Because Baidu had processed and combined text and images through AI, it was at fault. It ordered a written apology letter, refused a pinned apology on the website and rejected unproven damages claims. On 12 March 2026 the Nanjing Intermediate People's Court dismissed Baidu's appeal, finding a written apology adequate given the content's sporadic nature, the early stage of AI development and Baidu's later remedial steps.

Why it matters

Read together with the Hangzhou Internet Court's hallucination ruling, the case shows Chinese courts applying fault-based liability in both disputes, with the outcome turning on the nature of the content. Where output targets an identifiable person and lowers his reputation, the argument that hallucinations are technically unavoidable did not excuse the operator. In setting the remedy, the courts treated the stage of the technology and the operator's later remedial steps as mitigating. According to the reports, Baidu had not complied by May 2026 and the claimant applied for enforcement; a September 2026 report says he then received the written apology.

Related stages

The date of the first-instance judgment and the case numbers could not be verified. Much of the coverage originates with The Paper (澎湃新闻); Cover News (封面新闻) reported separately from judgment copies supplied by the claimant. The mitigating factors in the appeal reasoning appear only in The Paper's account. Reports say a similar suit by another lawyer against Baidu was filed with the Haidian District Court in Beijing in November 2025.

Similar decisions