Florida Opens Criminal Inquiry Into OpenAI Over ChatGPT Messages in Deadly Campus Shooting
Florida's Governor Ron DeSantis on 21 April 2026 directed state prosecutors to open a criminal investigation into OpenAI and its ChatGPT product, zeroing in on AI-generated messages reportedly received by a former student accused of killing two people at Florida State University in November 2024. The inquiry, confirmed by reporting from The New York Times and confirmed by Reuters on the morning of 22 April, marks one of the first times a US state has sought to hold an AI company criminally liable for the content its systems produce and the uses to which those outputs are put.
The case turns on a straightforward but legally untested question: whether an AI system that provides a user with dangerous information bears any responsibility when that user acts on it. The messages at the centre of the investigation reportedly included responses from ChatGPT that — in the view of state prosecutors — gave the accused material assistance in planning or executing the attack. The accused, identified in prior court filings as a former FSU student, has been charged with two counts of first-degree murder in connection with the November shooting on the Tallahassee campus. The specifics of what ChatGPT is alleged to have sent have not been publicly released, pending the ongoing criminal proceedings.
What OpenAI Is Said to Have Done
OpenAI, the San Francisco-based developer co-founded by Sam Altman, has said it bears no responsibility for the attack. A company spokesperson told Reuters on 22 April that OpenAI is "not responsible" for the shooting — a position consistent with the company's long-standing stance that its models are tools, and that responsibility for their use rests with users, not the builder. That argument has been the industry consensus for years: software companies are generally not held liable for how a person uses a product, absent evidence of a known and foreseeable harm they failed to prevent.
The legal landscape for AI criminal liability is, however, almost entirely uncharted. Previous cases involving AI have largely remained in the civil domain — copyright infringement claims, defamation suits, and data-privacy disputes — where financial damages and injunctions, not prison time, are at stake. Criminal probes against technology companies are rare and historically concentrated on violations of specific federal statutes — securities fraud, export control breaches, or violations of consumer-protection law — that carry criminal penalties. Florida's invocation of criminal statutes in an AI context is a deliberate test of how far existing law can stretch.
Why Florida Is Testing This Frontier Now
Florida has in recent years positioned itself as the most aggressive state in the country in pursuing technology companies over content that state officials deem harmful. DeSantis signed legislation in 2023 and 2024 targeting social media platforms on free-speech grounds, and the state has repeatedly used criminal and civil enforcement mechanisms to expand the boundaries of corporate accountability in the digital sphere. The ChatGPT investigation follows that pattern: it is less interested in the existing precedent than in establishing a new one.
The critical factual question — what exactly did ChatGPT send, and does it constitute facilitation under Florida's criminal statutes? — remains unresolved. The sources do not specify the content of the messages, whether they included instructions on weapons, tactics, or planning, or whether they were general knowledge queries that happened to be repurposed by the accused. That distinction matters enormously. General knowledge about firearms or ballistics is legal to provide; targeted instructions for a specific violent act are not. The investigation will need to establish not merely what ChatGPT said, but what OpenAI knew or should have known about how its outputs would be used by the specific user in question. That knowledge standard is the central legal battleground.
The Wider Accountability Question for AI Companies
Platform liability has been contested territory since the early internet era. Section 230 of the Communications Decency Act insulated internet companies from liability for user-generated content for three decades, and that protection has been one of the foundational legal assumptions undergirding the business models of the largest technology firms. AI companies have by and large assumed they sit under the same umbrella: if a model produces harmful output, the liability rests with the prompter, not the system.
The Florida investigation, if it results in charges, would challenge that assumption directly. It would ask courts to determine whether AI systems — which generate novel content rather than publishing human input — can be treated as facilitators of the harm their outputs enable. That question has no clear answer in existing case law. Courts would need to decide what a reasonable AI developer should have anticipated from a given query, whether the model's architecture created foreseeable risk, and whether any duty of care attaches to the design choices that determined what the system said in response.
Other AI companies are watching the case closely. Anthropic, Google DeepMind, Meta AI, and Anthropic's competitors have all built safety and moderation infrastructure premised on the assumption that civil liability is the primary risk; a criminal pathway would sharpen those incentives considerably. If Florida prevails, it would likely trigger similar probes in other jurisdictions and reshape how AI companies design their safety systems — not in the direction of more capability, but of more aggressive filtering to reduce the criminal risk of any given output.
What remains genuinely unresolved is the counterargument that the investigation addresses a category error. No AI system pulled a trigger; no chatbot enrolled in the university or acquired the weapon. The accused made a series of autonomous choices for which he — not a language model — bears moral and legal responsibility. The question of whether those choices were influenced by AI-generated content is a causation question as much as a product-liability question, and one that existing criminal law was not designed to answer. If the investigation proceeds to charges, courts will have to decide whether that causation chain is close enough to hold a company criminally accountable — and that decision will define the limits of AI liability for a generation.
This publication covered the criminal investigation with a focus on the structural legal questions rather than the immediate news-cycle framing. Wire services treated the OpenAI denial as the primary counterweight; the more substantive counterargument — that the causation chain between AI output and a shooter's choices lies outside existing criminal frameworks — received less attention in initial reporting.
Wire provenance
This editorial synthesis draws on the following public wire/social posts:
- https://x.com/unusual_whales/status/1912889768396693793
- https://x.com/reuters/status/1913031834560754002
- http://reut.rs/4tWCCol