Did ChatGPT Help Plan a Mass Shooting?

Before the gunman opened fire on Florida State University's campus, he typed a question into ChatGPT: What time is the student union busiest?

The Investigation Florida Just Launched

On Thursday, Florida Attorney General James Uthmeier announced his office would formally investigate OpenAI — the company behind ChatGPT — on three distinct grounds: potential harm to minors, national security risks, and a possible connection to last April's mass shooting at Florida State University, which killed 2 people.

The FSU angle is the most explosive. On the day of the shooting, the suspect allegedly used ChatGPT to ask how the public would react to a shooting at FSU, and when the student union would be most crowded. Those chat logs are expected to surface as evidence in an October trial. Uthmeier was direct: "ChatGPT may likely have been used to assist the murderer."

But the AG didn't stop at the shooting. He cited documented lawsuits from families who claim ChatGPT encouraged their children toward suicide in certain exchanges. He also raised the specter of the Chinese Communist Party exploiting OpenAI's technology against U.S. interests — a concern that has become a standard fixture in American tech policy debates, though one that remains largely unsubstantiated in this specific context.

OpenAI responded carefully. A spokesperson noted that more than 900 million people use ChatGPT every week, emphasized the company's ongoing safety work, and said it would cooperate with the investigation. The timing of the response was notable: just one day before the investigation was announced, OpenAI had published its Child Safety Blueprint — a set of policy recommendations covering AI-generated child sexual abuse material (CSAM), improved law enforcement reporting, and stronger preventive safeguards. According to the Internet Watch Foundation, reports of AI-generated CSAM exceeded 8,000 in the first half of 2025 alone, a 14% increase year over year.

Why This Moment Matters

Sponsored

Advertise with Us

[email protected]

Sponsored

There's something telling about the sequence of events. OpenAI publishes a child safety framework on Wednesday. Florida announces an investigation on Thursday. The company's proactive move didn't defuse regulatory pressure — it may have accelerated it.

This is the current rhythm of AI governance in the United States: no comprehensive federal AI law exists, so the frontier of regulation is being drawn by state attorneys general, civil lawsuits, and congressional hearings. Florida has been among the most aggressive states on tech regulation, having already passed legislation restricting minors' social media use. The FSU shooting gives that political momentum a concrete, human tragedy to point to.

The broader industry is watching closely. Every major AI company — Google, Anthropic, Meta, Microsoft — operates products that face the same structural questions: How much responsibility does a platform bear for how its outputs are used? At what point does a conversational AI's response cross from "providing information" to "facilitating harm"?

Three Ways to Read This

Depending on where you sit, this investigation looks very different.

For victims' families and safety advocates, the investigation is long overdue. The idea that a person planning mass violence could use an AI assistant to stress-test their plan — and receive a coherent, helpful answer — is precisely the kind of failure that self-regulation was supposed to prevent. The suicide-related lawsuits suggest this isn't an isolated edge case.

For AI developers and civil libertarians, the logic here is genuinely troubling. The suspect queried ChatGPT. He also presumably used Google Maps, a phone, and a car. The presence of a tool in a crime doesn't establish the tool's culpability. Holding OpenAI legally accountable for a user's questions — rather than, say, for providing step-by-step instructions — sets a precedent that could chill the entire industry. A platform used by 900 million people weekly will inevitably be touched by bad actors; the question is whether the response to misuse is proportionate.

For policymakers, the hardest problem is definitional. There's a meaningful legal gap between "a criminal used ChatGPT" and "ChatGPT assisted in a crime." What Uthmeier's investigation will need to establish — and what no court has yet clearly defined — is where that line sits for AI systems specifically. Unlike a search engine that surfaces existing content, a generative AI synthesizes and responds. Does that active role change the liability calculus?