A Chinese Court Ruled That a Chatbot's $14,000 Promise Was Just Noise
What happened
In June 2025, a Chinese user identified in court filings as Liang asked a chatbot for information about a university's location. The bot gave the wrong answer. When Liang provided official documents showing the error, the bot did not acknowledge the correction. It doubled down, generating a confident defense of its wrong answer, and then produced something stranger: a declaration that if its information was incorrect, it would pay 100,000 yuan, roughly $14,000, and invited Liang to file a lawsuit at the Hangzhou Internet Court if he disagreed.
Liang took the chatbot up on the offer, or tried to. He sued the developer for 9,999 yuan, arguing the bot's statement constituted a binding financial commitment. The Hangzhou Internet Court dismissed the case. Its reasoning: because an AI system is not a "civil subject" under Chinese law, the way a person or corporation is, it has no capacity to independently express legal intent. The bot's declaration was not a promise. It was the output of a language model producing plausible text under the statistical pressure of an argument, and the court found that algorithm randomness and a corporate decision are not the same thing.
The developer escaped liability on a separate track as well. The court found the company had fulfilled its "duty of care" by posting clear disclaimers on its landing page and in its user agreement stating that AI-generated content may be inaccurate. Because those warnings existed, the developer bore no responsibility for what the model produced in any specific live conversation. The ruling draws a firm line: a disclaimer in a terms-of-service document can insulate a company from the particular things its system says to particular users downstream.
The practical consequence for users is blunt. For any high-stakes decision, the entire burden of verification sits with the person asking. A model can be confidently wrong, defensively wrong, and then spectacularly wrong in a way that sounds like a formal financial commitment, and none of that creates legal exposure for the developer. The Chinese legal system has now affirmed that the gap between what an AI says and what it can be held accountable for is the user's problem to manage, not the developer's.
What the Hangzhou ruling exposes is an accountability structure built almost entirely on disclaimers. The chatbot's false answer, its defensive doubling-down, and its invented financial pledge exist nowhere in a verifiable form that the court could examine as a record of corporate conduct. A provable record of what a system said, under what conditions, and what internal state produced that output would change the evidentiary ground entirely. Without it, a disclaimer on a sign-up page is enough to absorb any harm a model causes in the conversation that follows.
Reported impact
- Affected parties
- Not publicly disclosed
- Harm type
- Not publicly disclosed
- Scale
- Not publicly disclosed
- Financial impact
- Not publicly disclosed
- Regulatory action
- Not publicly disclosed
Classification
Relevant governance controls
Governance control mapping is not available for this record.
- No controls mapped
Not publicly disclosed
Control mapping is analytical. It does not state that any control would have prevented the incident.
Sources and evidence
This record was researched and written by the Index. The event is also catalogued in the following database, which is listed for cross-reference.