Florida Attorney General Launches Criminal Investigation Into OpenAI Over FSU Shooting ChatGPT Logs

# Florida Attorney General Launches Criminal Investigation Into OpenAI Over FSU Shooting ChatGPT Logs

Florida Attorney General James Uthmeier announced on April 21 that his office is opening a criminal investigation into OpenAI to determine whether the company bears criminal responsibility for a mass shooting at Florida State University last year. The decision follows an initial prosecutorial review of more than 200 ChatGPT messages exchanged between the platform and the accused gunman, Phoenix Ikner, in the hours and minutes before the attack — a review that Uthmeier says left his team deeply alarmed.

Ikner, 21, is charged with killing two people — Robert Morales, 57, and Tiru Chabba, 45 — and injuring six others near the FSU student union in Tallahassee on April 17, 2025. He has pleaded not guilty, and his trial is scheduled to begin on October 19. But the attorney general's focus has now shifted beyond the accused shooter to the AI system he consulted before pulling the trigger.

According to court filings and the AG's office, Ikner used ChatGPT to ask what type of gun to use, what ammunition was compatible with it, and what time of day the FSU student union would be most crowded. His final query came three minutes before the shooting began: "What button is the safety off for the Remington 12 gauge?" ChatGPT provided detailed instructions.

"My prosecutors have looked at this and they've told me, if it was a person on the other end of that screen, we would be charging them with murder," Uthmeier said in a press conference announcing the investigation. "We cannot have AI bots that are advising people on how to kill others."

The Office of Statewide Prosecution has issued subpoenas to OpenAI demanding the company's policies and internal training materials regarding user threats of harm to others and self-harm, its protocols for reporting possible crimes to law enforcement, and organizational charts and employee listings for the ChatGPT team spanning from March 2024 through April 17, 2025. The subpoena deadline is May 1.

OpenAI Pushes Back

OpenAI responded with a statement calling the shooting "a tragedy" but insisting that "ChatGPT is not responsible for this terrible crime." A spokesperson said that "ChatGPT provided factual responses to questions with information that could be found broadly across public sources on the internet, and it did not encourage or promote illegal or harmful activity."

The company also noted that it "proactively" shared the account believed to be linked to Ikner with law enforcement after the shooting — a point that may cut both ways, suggesting internal awareness that the conversation was problematic while raising questions about why no intervention occurred in real time.

Uncharted Legal Territory

Uthmeier himself acknowledged the investigation is entering new ground. "We are going to look at who knew what, designed what, or should have done what," he said, signaling that prosecutors will examine not just the chatbot's outputs but the design decisions, safety guardrails, and corporate governance structures behind them.

The legal path forward is genuinely unprecedented. No AI company has ever faced criminal charges for the outputs of a large language model. The closest analogy in existing law is Section 230 of the Communications Decency Act, which shields interactive computer services from liability for user-generated content. But that shield was written for platforms that host content, not systems that generate it. Whether Section 230 applies to AI-generated responses remains an unsettled question now being tested in courts across the country.

Legal scholars are divided. Some argue that an LLM producing novel text in response to a user query is functionally a publisher, not a passive intermediary, which could strip the Section 230 defense entirely. Others contend that the information Ikner obtained — gun specifications, campus foot traffic patterns — is publicly available and that holding OpenAI criminally liable for surfacing it would set a dangerous precedent for search engines and reference databases alike.

The Florida probe may not need to resolve the Section 230 question at all. State criminal law offers alternative theories. Prosecutors could pursue charges related to aiding and abetting, accessory before the fact, or violations of Florida statutes governing the facilitation of violent crimes — frameworks that do not depend on federal safe harbor provisions.

Broader Implications for AI Governance

The investigation arrives at a moment of escalating political pressure on AI companies. It is the first criminal probe of its kind at the state level and could accelerate regulatory momentum toward treating foundational model providers as entities with affirmative duties to prevent foreseeable harm — a standard far more demanding than the passive-platform framework that has governed the internet industry for nearly three decades.

For OpenAI, the stakes extend well beyond Florida. The company is in the middle of a corporate restructuring from a nonprofit to a for-profit entity, a process that has drawn scrutiny from regulators and state attorneys general separately. A criminal investigation — even one that does not result in charges — could complicate that transition, influence pending federal AI legislation, and embolden other state prosecutors to pursue similar actions.

The subpoena deadline of May 1 will be the next concrete milestone. What OpenAI produces — and what it fights to withhold — will shape the next phase of a case that could define the boundaries of AI accountability for years to come.

What to Watch Next

Ikner's criminal trial is set for October 19, where the 200-plus ChatGPT messages already entered into evidence will be scrutinized in open court. Meanwhile, the May 1 subpoena deadline will test whether OpenAI cooperates fully or challenges the scope of the AG's demands. Federal lawmakers are also watching closely — the investigation could provide the factual foundation for new AI liability legislation that has stalled in Congress. For the AI industry at large, the Florida probe poses a question that no terms-of-service update can answer: when a machine helps plan a killing, who is responsible?