I Asked AI to Choose My Mother’s Burial Date. Then a Relative Was in a Car Crash. I Sued the Chatbot.

When AI says April 19 is fine: the legal ledger of the Doubao auspicious date case

On April 17, Mr. Shi’s mother died. He asked Doubao, an AI assistant, which day would be suitable for the burial. Doubao said April 19. Mr. Shi believed it. He notified relatives and arranged the funeral. Later he asked Doubao what time on that day would be suitable. Doubao answered between 7 a.m. and 9 a.m., but this time the answer also contained a sentence: April 19 is not a suitable day. Mr. Shi was confused. He asked several more times. Doubao said April 19 was not suitable. The relatives had already been informed. It was too late to change the date. He went ahead with the funeral. Soon after, a relative was seriously injured in a car accident. The relative blamed Mr. Shi. Mr. Shi complained to Doubao’s company but received no reply. He decided to sue. The complaint was written by Doubao.

The case was heard by the Jiashan County People’s Court. Doubao’s company said the user agreement states that generated content is for reference only and is not professional advice. Mr. Shi said that was a standard-form clause with insufficient warning.

The legal question is whether a platform must pay when its AI says something wrong.

1. Why product liability does not apply

The first question is which liability rule applies. One might think an AI is a product. If a product is defective, product liability applies. Product liability is strict liability. The plaintiff does not have to prove the manufacturer was at fault, only that the product was defective, that damage occurred, and that causation exists. That sounds favorable to users. But Doubao is not a product.

In September 2026, the Supreme People’s Court issued its Opinions on the Trial of Cases Involving Artificial Intelligence. Article 9 states that product liability applies only to AI products with a physical carrier. Pure software and cloud services are not subject to product liability. Doubao is a generative AI software service. It is a service, not a product. Article 3 of the Opinions also provides that legal liability for infringement caused by the use of AI shall be determined under the fault-liability principle in Article 1165(1) of the Civil Code. Mr. Shi must therefore prove that Doubao’s company was at fault.

The Hangzhou Internet Court reached the same conclusion in January 2026 in China’s first generative AI “hallucination” infringement case. The plaintiff, Liang, used an AI application to check university application information. The AI gave wrong information and promised in the conversation that “if the generated content is wrong, I will pay 100,000 yuan in compensation.” It suggested that Liang go to the Hangzhou Internet Court to sue. Liang did. The court dismissed all claims. The court gave four reasons why strict product liability did not apply. Generative AI lacks a specific, defined use and feasible quality-inspection standards. The generated information itself is not usually highly dangerous. The service provider lacks sufficient ability to foresee and control the generated content. Strict liability could improperly increase the provider’s burden and restrict the development of the AI industry.

Those four reasons apply to the Doubao case. There is no scientific or objective standard for an “auspicious burial date.” Doubao could not foresee that Mr. Shi would arrange a funeral based on its answer, still less that a car accident would happen afterward. Requiring Doubao to guarantee the accuracy of every generated sentence is technically impossible and commercially destructive. Mr. Shi cannot use product liability. He must prove fault.

2. What fault means: duty of care and the user agreement

The key question in fault liability is whether the platform fulfilled its duty of care. In the Hangzhou case, the court divided that duty into three layers. The provider must prominently explain the service function. It must ensure basic functional reliability. It must meet the contractual duty of care in the service agreement. If the provider fulfills these three layers, the court may find no fault.

Doubao’s user agreement states: “Generated content is for reference only and is not professional advice. For situations with major impact, it is recommended to consult relevant professionals. It should not be used as a basis for action or inaction. Any judgment or subsequent operation based on the output, and the consequences and responsibilities arising therefrom, shall be borne by the user.”

That is a prominent warning. It tells users not to treat AI words as professional opinion. For major matters, consult a professional. If you act on AI words, you bear the consequences. Mr. Shi asked about a burial date. That belongs to folklore and feng shui. That field has no unified standard, no authoritative certification, and no scientific method of verification. Doubao recommended April 19 and then changed its answer. That contradiction shows the instability of AI-generated content. The phrase “for reference only” warns about exactly that instability.

Mr. Shi argues that the user agreement is a standard-form clause with insufficient warning. That issue is discussed below. For fault, the court will ask whether the warning was prominent enough. Could users see it in the actual interface? Did the platform take reasonable technical measures to improve accuracy? If Doubao’s company can show it placed the agreement where users could easily see it, added warnings in the interface, and adopted error-prevention measures matching current technology, the court is likely to find no fault.

3. Causation: no legal line between the funeral date and the car accident

Even if Doubao’s company was at fault, Mr. Shi must prove a legal causal link between that fault and his relative’s car accident. Causation in tort law is not mere sequence in time. The fact that Doubao named a date and a car accident later happened does not establish causation. Legal causation requires a sufficient and foreseeable connection between the act and the harm.

Doubao recommended April 19 for burial. That is an output of information. The relative suffered a traffic accident. A traffic accident has its own causes: the risks of driving and traffic. The choice of burial date and the car accident do not have a cause-and-effect relationship. If such an indirect link counted as legal causation, everyone present that day would be liable. If someone had chatted with the injured relative a few more sentences, making him leave a few minutes earlier or later, the accident might not have happened. Push the logic further. The deceased also has causation. If the funeral had been a few days later, the accident might not have happened. That logic can extend forever, until everyone becomes a defendant.

The lawyer said Mr. Shi must prove fault and causation between the fault and the actual damage. That burden is very difficult. A mere wrong answer from an AI is clearly not enough. The Hangzhou Internet Court also held that AI has no civil subject status and cannot make expressions of intent. An AI’s self-generated “compensation promise” cannot be treated as the service provider’s expression of intent. By the same reasoning, an AI’s recommended “auspicious date” cannot be treated as Doubao’s promise to Mr. Shi. Mr. Shi believed it. That was his own choice.

4. Standard-form clauses: even if invalid, they do not save the case

Mr. Shi argues that the user agreement is a standard-form clause that increases user responsibility and lacks sufficient warning. That argument has a legal basis. Articles 496 and 497 of the Civil Code provide that the party providing a standard-form clause owes a duty to prompt and explain. A standard-form clause that excludes the other party’s main rights is invalid. AI service providers’ exemption clauses are subject to the negative list for standard-form clauses. If the platform failed to fulfill its prompting duty, the exemption clause may be invalidated.

The China Consumers Association has also stated that for AI customer service, important replies involving price, fees, and after-sales service cannot be simply exempted on the ground that they are automatically generated by algorithm. But the Doubao case differs from consumer-service scenarios. Mr. Shi was not transacting with Doubao’s company. He did not buy a “choose an auspicious date” service. He used Doubao as an information-query tool. Even if the exemption clause in the user agreement were invalidated, that would affect only the contractual allocation of responsibility. It would not change the liability rule or causation. Even if the court said a disclaimer was invalid, Mr. Shi would still have to prove independently that Doubao’s company was at fault and that the fault caused the car accident. Both are difficult.

5. Judicial policy: why AI does not carry the blame

The outcome is also shaped by judicial policy. The Supreme People’s Court’s Opinions adopt fault liability as the general rule. The policy consideration is to balance rights protection and industrial development, and to avoid imposing on service providers a burden disproportionate to their technical capacity. The Hangzhou Internet Court said strict liability could improperly increase providers’ burden and restrict the development of generative AI. That position is not unique to China. The European Union, in drafting its AI Liability Directive and revising the Product Liability Directive, has also struggled to balance innovation and victim protection.

The “hallucination” phenomenon of large models is an inherent feature of their technical architecture, not a defect that can be completely eliminated. If AI were required to guarantee the accuracy of every generated sentence, platforms could only become excessively conservative, limiting AI’s use in information queries and decision support. That does not mean AI providers can escape all responsibility. Platforms still must fulfill duties of prominent warning, technical error prevention, and content review proportionate to their risk-control capacity.

6. Back to that complaint

Mr. Shi’s case has been heard. The judgment has not yet come out. Under the elements of tort liability, his claim is unlikely to succeed. Tort liability requires strict elements: fault, damage, and causation, all of which must be proved. An AI giving a wrong answer is not enough to meet those elements. Attributing a car accident to a burial date chosen several days earlier based on an AI suggestion does not stand on the legal chain of causation.

One detail in this case matters more than the legal arguments. Mr. Shi decided to sue Doubao. The complaint was written by Doubao. In the conversation, Doubao told him he could sue and gave him the steps. The words on that paper were generated by AI. They now sit in the case file of the Jiashan County People’s Court.