OpenAI's state probe lands exactly where ChatGPT meets consumers

A coalition of U.S. state attorneys general is investigating OpenAI, according to TechCrunch and The Wall Street Journal, with New York's attorney general serving a subpoena. The questions reach beyond model safety into advertising, engagement, health data, minors and seniors.

The subpoena asks about ads, retention and vulnerable users

TechCrunch, citing The Wall Street Journal, reports that OpenAI received a subpoena from New York's attorney general as part of a broader investigation by a coalition of state attorneys general. The participating states have not been publicly specified.

The requested documents reportedly cover advertising, user engagement and retention, model sycophancy, consumer data, health data and the treatment of minors and seniors. OpenAI said it takes the concerns seriously and intends to engage constructively with the offices involved.

The company also pointed to protections for minors and people in difficult situations: age prediction, parental tools, restrictions on advertising aimed at children and safeguards that direct users toward human contacts or real world resources.

Consumer protection is moving faster than federal AI law

For product teams, the important part is that this does not look like an abstract frontier AI debate. It is conventional consumer protection logic applied to a chatbot that receives sensitive questions, shapes habits and can sound persuasive when it should slow down.

This cuts against the assumption that AI regulation will mostly arrive through a grand federal framework. State attorneys general already have tools for advertising, consumer protection, data and vulnerable groups. ChatGPT does not fit cleanly inside those categories, but it does not sit outside them either.

The risk is not only hallucination, but business design

Sycophancy is a technical problem, but engagement and retention are product decisions. If regulators start connecting safety failures to incentives for longer use, OpenAI will need to explain not only model behavior, but also the metrics the company optimizes around it.

The available reporting does not show a specific legal violation or formal accusation. This is a fact finding stage. Still, the scope is a warning to every company selling AI assistants as advisers, tutors, therapists or health companions without comparable responsibility.

Metrics and protections for weaker users will decide the outcome

The next signal is whether the investigation ends with no action, a settlement or a formal enforcement action. The most sensitive material will likely be internal documents on children, health data, self harm escalation and retention metrics.

The timing is also awkward for OpenAI. TechCrunch says the investigation arrives shortly after the company filed confidentially to go public. Public markets can tolerate risk, but they prefer to see it named in a prospectus, not pulled out by subpoena on the roadshow.

Lilith's verdict

ChatGPT is no longer standing only in front of a benchmark. It is facing a state attorney general with a binder of questions, and that person will ask who profits when a vulnerable user stays in the conversation 20 minutes longer.