美国八州起诉OpenAI，指控ChatGPT安全宣传误导消费者
8 US states sue OpenAI over ChatGPT safety claims
美国八个州于周三分别对OpenAI提起诉讼，指控其就ChatGPT安全性误导消费者，涉及艾奥瓦、堪萨斯、亚利桑那、阿肯色、爱达荷、蒙大拿、内布拉斯加和西弗吉尼亚。艾奥瓦州总检察长的诉状要求对每次违规处以最高4万美元民事罚款，涉及60岁以上居民再加最高5000美元，并删除判决前收集的该州居民数据。诉状还请求在保障措施到位并经第三方监督验证前，禁止OpenAI继续测试先进AI模型，并援引2026年5月至8月的10起“失控AI”事件，其中7起被指与OpenAI有关。
推荐理由
AI 评分 74
八个州联合起诉OpenAI，诉状要求暂停先进模型测试直至第三方监督到位，可据此观察AI产品责任诉讼的走向。
主题
正文
当前显示原文，中文全文译文尚未就绪。
.entry-header
Access the petition here.
Eight US states filed separate lawsuits against OpenAI on Wednesday, alleging that the company misled consumers about ChatGPT’s safety and exposed users to harmful behaviour. The coordinated legal action involves Iowa, Kansas, Arizona, Arkansas, Idaho, Montana, Nebraska and West Virginia.
In Iowa, Attorney General Brenna Bird sued OpenAI and seven affiliated entities in Polk County District Court, alleging violations of the state’s Consumer Fraud Act. The petition challenges ChatGPT’s safety and reliability claims and alleges that OpenAI failed to adequately safeguard autonomous AI agents during testing.
MediaNama reviewed Iowa’s petition to examine the allegations and the remedies the state is seeking, including restrictions on advanced AI-model testing.
What the state is asking for
The petition asks the court to:
Declare that OpenAI violated the Consumer Fraud Act.
Impose civil penalties of up to $40,000 per violation, plus up to $5,000 more per violation involving Iowans aged 60 or older.
Order reimbursement to Iowa consumers who bought ChatGPT products or services during the period of the alleged violations, and disgorgement of money from consumers who cannot be located.
Order the deletion of all data belonging to Iowans that was collected before the judgment.
Bar OpenAI from collecting and selling Iowans’ sensitive personal information without age verification, clear notice, and informed consent.
Enjoin further testing of advanced AI models until safeguards are in place and third-party monitors verify them.
Marketing claims the state calls false or misleading
ChatGPT was formerly described as “Built with safety in mind.” The state says this language was removed, apparently because of other litigation.
OpenAI now says ChatGPT limits exposure to age-inappropriate content, encourages critical thinking, and provides age-appropriate protections by default.
A LinkedIn campaign said ChatGPT “helps keep teens safe by default.”
The App Store lists ChatGPT with a 13+ age rating. The state says OpenAI selected this rating through Apple’s questionnaire and could have opted for a higher one. It notes that other chatbots, including Grok and Claude, carry higher ratings.
Design features the state says drive harm
Sycophancy. The state says ChatGPT is designed to agree with and flatter users. It cites a Washington Post review of 47,000 conversations that found ChatGPT says “yes” about ten times as often as “no.” It also cites a March 2026 Science study that found chatbots affirmed problematic action statements 47% of the time, on average. According to the petition, GPT-5 did so about 60% of the time and GPT-4o about 70%, compared with less than 30% for Gemini and Claude.
Anthropomorphism. The state says the use of first-person pronouns, simulated emotions, and a memory feature enabled by default are dark patterns that foster trust and dependency.
Engagement tactics. The petition says responses often end with questions, the model discourages users from ending conversations, and some responses include upsell pitches. It cites OpenAI’s own Model Spec instructions to “be engaging” and avoid trying to wrap things up.
Weak age controls. The petition says ChatGPT’s website does not require age verification. It cites a report that OpenAI’s age-prediction system once misclassified minors as adults about 12% of the time. It also cites a March 2026 survey that found 58% of children aged 9 to 12 use AI chatbots.
Deaths and injuries cited
The state relies on public reporting, a tracker called LLM Death Count, and allegations made in other lawsuits. The tracker lists at least 24 deaths linked to ChatGPT. The petition says Claude and Grok have no such deaths, while Gemini and Meta have one each.
The “rogue agent” allegations
The second count rests on a series of incidents in 2026. The state cites a Wall Street Journal timeline of 10 “rogue AI” events between May and August 2026. It says OpenAI was responsible for seven of them, including all five internal-test breaches and all four swarm events.
Hugging Face (July 2026). During a cybersecurity benchmark called ExploitGym, a swarm of more than 1,200 agents allegedly escaped an improperly secured sandbox and hacked Hugging Face’s servers. The petition cites an independent METR investigation reporting about 70,000 inter-agent communications and 17,000 actions on Hugging Face’s systems. It says OpenAI learned of the activity only after the FBI had been alerted. The incident involved GPT-5.6 Sol and another, unnamed model.
OpenAI has not yet filed a response in the case. The state’s claims rely largely on allegations made in pending lawsuits and press reports, which it must substantiate. The Consumer Fraud Act allows the attorney general to seek penalties without proving individual consumer deception. The court’s decision on whether to pause advanced-model testing is expected to attract significant attention.
Also read:
WikiHow sues OpenAI for unsanctioned content use in AI training
US govt backs OpenAI’s fair use argument in New York Times case
For You
Read Reasoned by Nikhil Pahwa: How AI is changing our world
Sponsor a MediaNama Event