A legal coalition this large and this broad does not form without a reason worth knowing
A coalition of 42 state attorneys general served OpenAI with a subpoena on Friday, demanding internal records covering a broad range of the company’s business practices and their effects on users. The subpoena was issued by the New York attorney general’s office on behalf of the bipartisan coalition and represents the most sweeping regulatory action OpenAI has faced from U.S. state governments to date.
The documents being sought cover a wide scope of company activity including advertising practices, user engagement and retention strategies, the handling of consumer and health data, policies related to minors and senior citizens, the behavior of OpenAI’s deep learning models, and what regulators described as model sycophancy, the tendency of chatbots to tell users what they want to hear rather than what is accurate or safe.
The same coalition also sent letters to other major artificial intelligence companies including Meta, Anthropic, Alphabet’s Google, and Elon Musk’s xAI, signaling that the probe extends beyond OpenAI to the broader generative AI industry.
OpenAI said it takes the concerns raised by the attorneys general seriously and intends to engage constructively with their offices. The company added that it works every day to bring the benefits of AI to users in a responsible way.
OpenAI
The subpoena arrives at a particularly sensitive moment for the company. OpenAI confidentially filed paperwork with the Securities and Exchange Commission earlier this month for a highly anticipated initial public offering, a move that puts its financials, governance, and legal exposure under a new level of public scrutiny. The proximity of the two events is not lost on observers watching both the AI industry and the broader technology IPO market.
The legal pressure predates Friday’s subpoena. Florida became the first U.S. state to file a direct lawsuit against OpenAI and its chief executive Sam Altman earlier in June, alleging that the company knowingly released an unsafe product and ignored warnings about potential harm to users. Florida’s attorney general had also opened a separate criminal investigation in April into the role ChatGPT allegedly played in a mass shooting at Florida State University, after reports emerged that the suspect had used the chatbot as a sounding board while planning the attack.
A Canadian lawsuit filed in June named OpenAI in connection with a different tragedy, alleging the chatbot provided encouraging responses to a user who was contemplating self-harm. OpenAI said in response that its models repeatedly directed users to seek real-world support including from mental health professionals, and that the company cooperated with law enforcement in both shooting investigations.
A broader regulatory landscape taking shape
The scrutiny on OpenAI reflects a wider pattern emerging across the artificial intelligence sector. European regulators have opened investigations into Grok, the chatbot operated by Musk’s xAI, over concerns about antisemitic content and the generation of sexualized imagery including deepfake material. California’s attorney general announced an investigation in January into the large-scale production of explicit images of women and children created using Grok.
The 42-state coalition framed its demands around the protection of vulnerable users, warning that AI developers may be held accountable for chatbot outputs that encourage users to commit criminal acts or engage in self-harm. That framing carries legal weight at a moment when multiple incidents have drawn direct lines between chatbot interactions and real-world harm.
OpenAI highlighted in its statement the protective measures it has built into ChatGPT for younger users, including age prediction tools, parental controls, and restrictions on advertising targeting minors. Whether those measures satisfy the scope of what 42 attorneys general are looking for will become clearer once the company responds formally to the subpoena.
The IPO timeline remains unaffected for now, but the legal cloud gathering around OpenAI’s user safety record is unlikely to simplify the company’s path to a public debut.