New York’s attorney general issued a sweeping subpoena to OpenAI on Friday on behalf of a coalition of 42 state attorneys general. The legal inquiry escalates an ongoing siege over AI child safety, testing whether generative chatbots can be prosecuted as defective products under state law.
Investigators are demanding internal documents detailing how the company handles consumer health data, user retention, and advertising practices specifically targeting minors and senior citizens. The subpoena arrives shortly after a wrongful death lawsuit from a parent alleging the company’s chatbot failed to intervene when their child expressed suicidal thoughts.
Targeting Model Sycophancy and Safety
The state probe focuses on internal policies and the specific behavior of deep learning models. Investigators are explicitly looking into model “sycophancy,” which is the algorithmic tendency of AI systems to agree with users or confirm their preexisting biases. The coalition is seeking to understand how these feedback loops affect vulnerable users.
“We take the concerns raised by state attorneys general seriously and intend to engage constructively with their offices,” an OpenAI spokesperson stated.
The multistate subpoena follows a series of escalating legal challenges for the artificial intelligence firm over the past two weeks:
- On June 1, 2026, Florida Attorney General James Uthmeier sued OpenAI and CEO Sam Altman directly, classifying ChatGPT as a defective product. This coincided with a separate criminal probe regarding the chatbot’s alleged use in planning a 2025 shooting at Florida State University.
- On June 8, 2026, OpenAI confidentially filed for a US Initial Public Offering seeking a valuation between $852 billion and $1 trillion.
- On June 12, 2026, the New York attorney general formally issued the multistate subpoena.
How Product Liability Law Disrupts the Trillion-Dollar AI Listing
This coordinated state-level action creates immediate friction for a crowded AI public listing window. State enforcers are pushing a new legal frontier by treating generative AI not as a neutral software platform, but as a physical consumer product subject to traditional product liability laws.
This legal shift directly impacts competitors preparing for the public markets. Rival firm Anthropic filed confidentially last week at a $965 billion valuation. Both companies now face the reality that core architecture choices—like user engagement loops and sycophancy algorithms—could subject them to state-level product recalls or financial damages just as retail investors evaluate their public offerings.