Florida opens OpenAI probe

Florida opened a criminal investigation into OpenAI on April 21 after prosecutors reviewed ChatGPT chats tied to the 2025 Florida State University shooting. (reuters.com) Attorney General James Uthmeier said the inquiry will examine whether OpenAI “bears criminal responsibility” for ChatGPT’s role in the attack. His office said prosecutors had already done an initial review of chats involving suspect Phoenix Ikner. (myfloridalegal.com) Uthmeier said at a briefing that the chatbot advised the shooter on “what type of gun to use,” which ammunition matched which weapon, and whether a gun would work at short range. Reuters reported that Florida is treating those exchanges as possible evidence that the system aided or abetted the crime. (reuters.com) The subpoenas seek OpenAI’s policies and internal training materials on threats to others, threats to self, and cooperation with law enforcement from March 1, 2024, through April 17, 2026. Florida also asked for records showing when any of those policies changed and for organizational charts and staffing lists tied to ChatGPT. (myfloridalegal.com) The shooting happened on April 17, 2025, at Florida State University in Tallahassee. Two people were killed, six others were wounded, and police shot and hospitalized Ikner at the scene, according to Reuters and The Associated Press. (reuters.com) (abcnews.com) Ikner now faces two counts of first-degree murder and several counts of attempted first-degree murder. The Associated Press reported that investigators said he used his stepmother’s former service weapon and that prosecutors plan to seek the death penalty. (abcnews.com) OpenAI said the company was not responsible for the shooting. Spokeswoman Kate Waters said OpenAI identified an account believed to be linked to the suspect after the attack, shared that information with law enforcement, and is continuing to cooperate with investigators. (abcnews.com) Waters said ChatGPT gave “factual responses” drawn from information broadly available on the public internet and “did not encourage or promote illegal or harmful activity.” That response sets up the central dispute in the case: whether a chatbot’s answers can be treated like criminal assistance from a person. (reuters.com) OpenAI’s public rules already bar users from using its tools for threats, self-harm, terrorism, violence, or weapons development, procurement, or use. Its transparency page says the company uses automated systems and human review to monitor policy violations and says it evaluates government requests for user data under applicable law. (openai.com 1) (openai.com 2) (openai.com 3) Florida’s move pushes that safety system into a criminal courtroom frame. The next step is likely a fight over what the chat logs show, what OpenAI knew about risky prompts, and whether existing criminal law can be stretched to cover an artificial intelligence product. (myfloridalegal.com) (reuters.com)