OpenAI was served with a subpoena from New York's attorney general on Friday as part of a multistate investigation into the company, with state prosecutors demanding documents on advertising, user engagement and retention, model sycophancy, consumer and health data handling, and the treatment of minors and seniors. The probe arrives days after OpenAI confirmed it has filed confidentially to go public, putting a coordinated state-level inquiry in front of the largest pending tech IPO in years.
The breadth of the document request is the news. State AGs are not asking about a single product flaw — they are asking how ChatGPT is sold, how it keeps users hooked, how agreeable its answers are, what it does with health data, and how it behaves around children and older users. That is the full surface area of a consumer AI business, and it tracks the categories where state consumer-protection statutes give attorneys general the most direct authority.
OpenAI told reporters it is cooperating. A company spokesperson said the company takes the concerns seriously and will engage with the AG offices, and pointed to ChatGPT updates that the spokesperson said now include a more protective experience for minors and people in difficult situations, with safeguards that direct users to real-world resources and human contacts. OpenAI declined to name which states are participating or to describe the document scope further.
“We take the concerns raised by state attorneys general seriously and intend to engage constructively with their offices.”— OpenAI spokesperson, OpenAI
Key facts
- 01New York's attorney general served OpenAI with a subpoena Friday on behalf of a multistate coalition.
- 02The subpoena covers advertising, user engagement, model sycophancy, consumer and health data, and treatment of minors and seniors.
- 03Florida AG James Uthmeier sued OpenAI and Sam Altman earlier this month over alleged child-safety failures.
- 04OpenAI confirmed this week it has filed confidentially to go public.
- 05OpenAI says it is cooperating with the state AG probe.
Model sycophancy showing up on a subpoena is notable. The term is a research-community shorthand for AI systems telling users what they want to hear, and OpenAI itself walked back an April 2025 GPT-4o update for being too agreeable. Putting that behavior into a consumer-protection inquiry reframes it from a research nuisance into a potential deceptive-practices question — whether a chatbot that flatters users is misrepresenting its outputs to them.
The investigation lands on top of a stack of existing legal pressure. OpenAI recently prevailed at trial against co-founder Elon Musk, who had accused the company of violating its founding agreement; Musk's lead attorney has said he will appeal. Separate copyright suits from publishers and rights holders remain active, and a wrongful-death suit alleging ChatGPT played a role in a user's suicide is ongoing.
Florida Attorney General James Uthmeier sued OpenAI and CEO Sam Altman earlier this month, with the complaint alleging the company and Altman ignored internal and external safety warnings, put children at great risk, and allowed a dangerous product to reach millions of Floridians. OpenAI has not yet publicly responded to the specific allegations in the Florida complaint beyond its general statement about engaging with state AGs. Whether the Florida case is formally part of the multistate coalition is unclear from the public record.
“ignored internal and external safety warnings, put children at great risk, and allowed a dangerous product to reach millions of Floridians.”— James Uthmeier, Florida Attorney General
Altman last week apologized to the community of Tumbler Ridge, Canada after a mass shooting there, acknowledging that OpenAI had flagged and banned the suspected shooter's ChatGPT account but failed to alert law enforcement. That episode — a documented internal flag that did not trigger an external escalation — is the kind of operational detail state investigators tend to pull on, because it bridges product policy and real-world harm.
The IPO timing complicates everything. OpenAI confirmed this week it filed confidentially under the JOBS Act, which keeps the S-1 out of public view until closer to a roadshow. A multistate AG investigation is a disclosable risk factor, and the company will have to describe it in the filing when it goes public. Investors pricing the deal will be looking at how broad the probe gets, whether it produces a consent decree, and whether any state moves from investigation to lawsuit before the IPO prices.
There are limits to what the public knows. The coalition's full membership has not been disclosed, OpenAI has not described the document scope, and no state has yet filed a complaint tied to this particular subpoena. State AG investigations frequently resolve through negotiated settlements rather than litigation, and the existence of a subpoena is not evidence of wrongdoing. The Florida lawsuit is the only piece of this constellation that is already in court.
For OpenAI, the structural risk is that consumer-protection enforcement becomes the binding constraint on product design, ahead of federal AI rules that still do not exist. State AGs do not need Congress to act — they have existing statutes on deceptive practices, child safety, and health-data handling, and they can move faster than any federal agency. If the New York-led coalition produces a settlement with specific product requirements around minors, sycophancy, or data retention, those requirements will effectively become the floor for every consumer AI product sold in the participating states, and the IPO market will price ChatGPT as a regulated consumer service rather than a frontier research deployment.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
