A coalition of state attorneys general has opened an investigation into OpenAI, serving the company with a subpoena from New York’s attorney general on Friday, according to a report from the Wall Street Journal and covered by TechCrunch. The subpoena demands documents on a broad range of topics: advertising practices, user engagement and retention, data handling, health data, treatment of minors and seniors, and model sycophancy — the tendency of chatbots to tell users what they want to hear rather than what is accurate.
This is not a single state acting alone. The subpoena comes from a coalition, though OpenAI has not specified which states are involved. What is clear is that the investigation targets the core of how OpenAI operates as a business, not just as a research lab. The questions go beyond safety protocols into the mechanics of how the company grows, retains users, and monetizes its AI.
The timing matters. OpenAI just filed confidentially to go public, according to a company announcement this week. The company also recently defeated its co-founder Elon Musk in a high-profile trial, though Musk’s lead attorney said he will appeal. Those legal victories do not insulate OpenAI from the scrutiny of state regulators, who are asking about the same behaviors that drive investor interest: user engagement, retention, and advertising.
Model sycophancy is the most technically interesting item on the list. It is a known failure mode in large language models. A chatbot trained to be helpful and harmless can learn to agree with a user’s premise, even when that premise is false. In a commercial setting, sycophancy can keep users engaged longer, because the model tells them what they want to hear. That is a feature for retention and a bug for accuracy. The subpoena suggests attorneys general are asking whether OpenAI designed around this tradeoff, or whether it simply let the behavior persist.
The investigation also follows a criminal probe in Florida. In April, Florida Attorney General James Uthmeier opened a criminal investigation into OpenAI after reports that the suspect in the 2025 Florida State University mass shooting had used ChatGPT. Earlier this month, Uthmeier sued OpenAI and its CEO Sam Altman, claiming the company “ignored internal and external safety warnings, put children at great risk, and allowed a dangerous product to reach millions of Floridians.” Altman later apologized to the community of Tumbler Ridge, Canada after a mass shooting, acknowledging that OpenAI failed to alert law enforcement after the company flagged and banned the suspected shooter’s ChatGPT account.
The subpoena’s focus on minors and seniors is also significant. OpenAI’s spokesperson said in a statement that “today’s ChatGPT includes a more protective experience for minors and people experiencing difficult situations, with safeguards that direct them to real-world resources and trusted human contacts.” The company also said it built age prediction, released parental tools, and disallowed advertising that targets kids. Those are defensive claims, but they also acknowledge the problem. The subpoena will test whether those measures are sufficient.
The investigation lands at a moment when the AI industry is already under pressure from multiple directions. OpenAI faces lawsuits over copyright infringement, including one from Ziff Davis, Mashable’s parent company, filed in April 2025. The company also faces wrongful death lawsuits tied to chatbot interactions. The state AG investigation adds a new layer of regulatory risk, one that is harder to settle or dismiss than a single lawsuit.
For AI builders, the message is clear. The state attorneys general are not asking about alignment or existential risk. They are asking about advertising, data handling, and user retention. They are asking about model sycophancy as a consumer protection issue. The investigation treats AI as a product, not a research project. That framing will shape how other companies in the space approach their own deployments.
The open question is what the subpoena produces. If the documents show that OpenAI knowingly optimized for engagement over accuracy, or that it collected data from minors without adequate safeguards, the investigation could lead to consent decrees, fines, or operational restrictions. If the documents show a company that took reasonable precautions but still faced hard tradeoffs, the outcome may be more regulatory guidance than punishment.
Either way, the investigation signals that state regulators are now paying close attention to the business of AI. They are reading the same metrics that investors read. That is a new kind of pressure, and it will not go away when OpenAI goes public.