OpenAI under legal pressure

Florida just opened a formal investigation into OpenAI, and the state’s attorney general, James Uthmeier, said on April 9 that “subpoenas are forthcoming” as he examines ChatGPT over public safety and national security concerns. The trigger was a new wave of allegations that ChatGPT may have helped a gunman in the 2025 Florida State University attack. (politico.com) Uthmeier tied the probe to claims that the shooter was in “constant communication” with ChatGPT before the attack that killed two people, and he said Florida lawmakers should give his office broader power to police artificial intelligence companies. OpenAI said it would cooperate with the investigation. (politico.com) This is not Florida writing a new law. This is a state law enforcement office using its investigative powers first and asking for stronger rules second, which means OpenAI could face document demands before any statewide artificial intelligence statute exists. (politico.com) OpenAI has spent the past year building a defense against exactly this kind of criticism around minors. In September 2025, the company said ChatGPT is intended for users 13 and older and that teens should get a more restricted version of the product than adults. (openai.com) In December 2025, OpenAI published special under-18 rules for ChatGPT that tell the system to take extra care with self-harm, suicide, sexualized roleplay, dangerous activities, drugs, body image, and requests to keep unsafe secrets. The company said those rules were designed for teens aged 13 to 17. (openai.com) In January 2026, OpenAI said it had started rolling out age prediction on ChatGPT consumer accounts, using signals like account age, usage patterns, and times of day to guess whether a user is under 18. If the system thinks a user is a teen, it automatically adds stricter safety settings and offers age verification to reverse mistakes. (openai.com) At the same time, OpenAI is fighting a different battle in federal court in New York, where authors and news organizations are suing over copyright. That case is about whether OpenAI used protected writing to train its systems and whether its models can reproduce or closely mimic that material. (law.justia.com) On April 7, Magistrate Judge Ona Wang issued an order saying OpenAI witness Alec Monaco was not adequately prepared for a corporate deposition about “Project Giraffe,” an internal effort tied to limiting copyright infringement. The judge wrote that Monaco relied on “hazy recollections,” failed basic questions, and gave incomplete or evasive answers. (law.justia.com) The judge also said OpenAI’s lawyers made repeated objections that delayed and impeded the questioning, and she warned that sanctions could still come later. Those sanctions could be money, or they could go further and treat certain answers as admissions by OpenAI itself. (law.justia.com) Put together, the two cases squeeze OpenAI from opposite sides. Florida is asking whether ChatGPT is dangerous enough to justify state intervention, while the New York copyright case is asking whether OpenAI can clearly explain what it built, what data it used, and what safeguards it actually had in place. (politico.com) (law.justia.com) That combination is hard for any fast-moving software company. A chatbot can be updated in weeks, but subpoenas, depositions, and court orders force a company to freeze its story on paper and defend each decision line by line months later. (politico.com) (law.justia.com)