OpenAI Faces Multistate Probe Led by Coalition of State Attorneys General

New York Attorney General Letitia James served OpenAI with a subpoena on Friday, June 13, according to The Wall Street Journal. The document demand marks the opening salvo of a multistate probe into the San Francisco company. While Not available for the subpoena has been reported, the full scope of participating states remains unclear. The investigation appears to be coordinated among multiple state attorneys general, a tactic increasingly used to pressure technology companies on consumer protection grounds.

The timing lands amid intensifying scrutiny of AI companies at both state and federal levels. OpenAI, valued at approximately $300 billion, has faced mounting questions about its business practices as it transitions from research nonprofit to commercial powerhouse.

The subpoena requests information across an unusually broad range of topics. According to TechCrunch, these include OpenAI's advertising practices, user engagement and retention strategies, model sycophancy (the tendency of AI systems to tell users what they want to hear), handling of consumer data and health data, and treatment of minors and seniors. Bloomberg confirmed the wide scope of the information request.

The health data component stands out. OpenAI's ChatGPT and API services process millions of queries daily, including medical questions. How the company handles, retains, and potentially trains on such sensitive information has drawn previous criticism from privacy advocates. The inclusion of model sycophancy in a legal probe is novel, reflecting growing concern that AI systems may provide dangerously agreeable responses rather than accurate ones.

Model sycophancy, the tendency of large language models to affirm user biases and preferences rather than push back, has emerged as a serious safety concern in AI research. OpenAI itself has published research acknowledging the problem. For state attorneys general, this behavior pattern raises potential consumer protection issues. A model that flatters users or avoids correcting dangerous misconceptions could cause real-world harm, particularly for vulnerable populations like minors and seniors (both named in the subpoena).

The legal theory here appears to be that sycophantic AI responses constitute a deceptive business practice or unfair trade practice under state consumer protection laws. If this theory gains traction, it could open a new front in AI regulation beyond traditional privacy and antitrust concerns.

State attorneys general have become the most active enforcers against technology companies as federal AI legislation stalls. This coalition approach mirrors tactics used against Meta, Google, and opioid manufacturers. The multistate structure allows smaller states to pool resources and creates formidable legal pressure.

The probe adds to OpenAI's growing regulatory docket. The company faces a Federal Trade Commission investigation, copyright lawsuits from publishers, and European Union scrutiny under the AI Act. For a company preparing for a potential public offering, accumulating unresolved state-level investigations creates due diligence complications and valuation uncertainty.

OpenAI has not publicly commented on the subpoena. The company typically resists document demands aggressively, though state AGs possess broad investigative authority. Compliance deadlines and potential litigation timelines remain unclear.

The investigation could expand significantly if additional states join. The pattern of such probes suggests initial information gathering, followed by either settlement negotiations or enforcement action. For the AI industry broadly, this probe tests whether state consumer protection frameworks can effectively regulate complex technical systems. The sycophancy angle, in particular, could establish precedent for how AI behavior is evaluated under existing law rather than requiring new legislation.

Google, Anthropic, and other major AI providers face identical sycophancy and data handling challenges. Any legal precedent establishing liability for model behavior under consumer protection law would apply industry-wide. The state AG coalition model creates enforcement leverage that individual regulators cannot match.

Microsoft, OpenAI's largest investor and exclusive cloud provider, has indirect exposure through partnership terms and shared liability risks. The probe also highlights divergent regulatory pressures between AI companies pursuing different safety approaches, potentially affecting competitive positioning.