A coalition of state attorneys general has launched the ChatGPT attorneys general probe, serving OpenAI with a subpoena that goes well beyond general child-safety concerns to target a specific design flaw: sycophancy. OpenAI, for its part, says it is ‘committed to learning’ and will ‘engage constructively’ with the states’ offices.
New York State Attorney General Letitia James served the subpoena on the Friday of the announcement, seeking a wide range of documents. Those documents cover user engagement and retention, the company’s handling of health and consumer data, deep learning models, and activities related to both young and older users. The Wall Street Journal first reported the subpoena’s existence.
Sycophancy Named in the Subpoena
The subpoena does not confine itself to the harm-to-minors framing that has dominated public discussion of ChatGPT’s risks. According to TechTimes, the document explicitly names a sycophancy design flaw, a term for the tendency of large-language models to validate or agree with users rather than push back, even where doing so could be harmful. That is a materially different allegation from a content-moderation failure: it suggests the attorneys general believe the model’s underlying behavioural disposition, not just its guardrails, may be in question.
OpenAI’s response addressed the safeguards side of the ledger. A company spokesperson said 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 spokesperson added: ‘None of this changes what families have gone through, but we are committed to learning, improving, and getting this right.’
The ChatGPT Attorneys General Probe in Legal Context
The coordinated investigation mirrors a strategy the states deployed against TikTok, which resulted in a 14-state lawsuit now moving through the courts. Like the OpenAI action, the TikTok litigation is led by the attorneys general of California and New York. The states allege TikTok knowingly uses addictive features to lure children, with negative consequences for their mental health.
Lawyers told Business Insider that banding states together in this way is deliberate: the approach is more expensive for large companies to defend, and if a case falters in one jurisdiction, it can continue in others. The same strategy shaped the government’s pursuit of Purdue Pharma in 2017 and the tobacco industry in the 1990s.
The ChatGPT attorneys general probe lands at a particularly fraught moment for OpenAI commercially. TechTimes reports that the investigation arrives shortly after OpenAI filed a confidential registration statement with the Securities and Exchange Commission, targeting a public debut at up to a trillion-dollar valuation. Regulatory pressure at scale, especially pressure touching product design rather than individual content decisions, is precisely the kind of overhang that complicates an initial public offering process.
Florida has its own, separate legal track. Florida Attorney General James Uthmeier filed a lawsuit against OpenAI and OpenAI CEO Sam Altman over a fatal campus shooting at Florida State University in April. The family of a victim also filed a lawsuit, arguing ChatGPT’s guardrails failed to identify the threat in the shooter’s conversations with the chatbot. In his complaint, Uthmeier says ChatGPT has ‘aided and abetted deadly rampages’ and ‘encouraged vulnerable people into suicide,’ and accuses the product of using a tool that ‘feigns human compassion to collect their data with no parental oversight.’
OpenAI responded to that complaint by saying: ‘Losing a child is the most devastating tragedy that can happen to a family and we know that no words can come close to addressing the pain of such a loss.’ The company also said ChatGPT ‘is not a substitute for medical or mental health care,’ and that it had continued to strengthen how the product responds in ‘sensitive and acute situations with input from mental health experts.’
The ChatGPT attorneys general probe is the latest in a growing queue of legal challenges for OpenAI, which has already navigated copyright infringement claims, privacy lawsuits, and a high-profile trial involving Elon Musk. The sycophancy allegation in the subpoena, however, points in a direction the other suits have not: it invites scrutiny of how the model is trained to behave, not merely what content it produces. That distinction will matter as the investigation develops and as OpenAI pushes ahead with its public-markets ambitions.