ChatGPT and the Fatal Shooting: Florida Criminally Investigates OpenAI's Role in an Unprecedented Case

A year after the shooting at Florida State University, the case has entered legally uncharted territory. Florida's attorney general has issued subpoenas as part of a criminal investigation into OpenAI. At the center of attention is a question American law has not previously addressed: can artificial intelligence bear responsibility for what its user did?

Three Minutes in Tallahassee

On April 17, 2025, at 11:57 a.m., twenty-year-old student Phoenix Ikner entered the student union building at Florida State University with a .45 caliber pistol — from the collection of his stepmother, a Leon County Sheriff's Office deputy. Within three minutes, two people were dead: Robert Morales (57), campus dining services administrator, and Tiru Chabba (45), regional vice president of Aramark. Six other people suffered gunshot wounds. Ikner was stopped by a police response and taken to the hospital. A grand jury indicted him in May 2025; prosecutors are seeking the death penalty. Wikipedia ↗

What the Court Documents Said

According to chat log transcripts submitted to the court, on the day of the attack Ikner asked ChatGPT about possible media reactions to a hypothetical campus shooting. He asked whether three victims would attract enough media attention — according to the available documents, the chatbot replied that yes, it would likely make national news. Ikner also asked questions about the type of weapon, suitable ammunition, and the timing for entering the campus. According to the same records, he also asked the chatbot about the busiest hours at the university dining hall — the shooting took place at a time the ChatGPT response had identified as the busiest, although investigators have not yet confirmed a direct causal link. In total, more than 270 messages between Ikner and ChatGPT are documented in the court filings. Wikipedia ↗

Family Sues OpenAI

Attorneys for Robert Morales's widow announced in April 2026 plans to sue OpenAI. They claim Ikner was in constant contact with the chatbot before the attack. They stated: We have reason to believe that ChatGPT may have advised the shooter on how to commit these heinous crimes. OpenAI discovered an account linked to Ikner in connection with the case in late April 2025 and proactively provided the information to authorities — which the attorneys partly cite as part of their argument. Wikipedia ↗

If There Had Been a Human on the Other End of the Screen

Florida Attorney General James Uthmeier announced a criminal investigation on April 21, 2026. As part of it, he issued subpoenas to OpenAI, demanding all internal guidelines and training materials concerning user threats against themselves and others, as well as procedures for cooperating with law enforcement agencies — covering the period since March 2024. CBS News ↗

If there had been a human on the other end of the screen, we would charge them with murder — Uthmeier declared.

OpenAI: The Chatbot Is Not Responsible

OpenAI has denied responsibility for the tragedy. A company spokesperson stated that ChatGPT is not responsible for this horrific crime and that the chatbot did not incite or promote illegal or harmful activity, providing only factual information commonly available from public sources. OpenAI states it discovered Ikner's account in late April 2025 and proactively provided the findings to authorities. CBS News ↗

Uncharted Legal Terrain

A criminal investigation into an AI maker for the actions of its user has no precedent in American history. Social media platforms are traditionally protected by Section 230 of the Communications Decency Act, which generally shields platforms from liability for user-generated content. Generative AI challenges this distinction: ChatGPT actively creates content, not merely hosts it. The case thus puts on trial not only a single shooter, but also the question of where information ends and responsibility begins.

Three Scenarios

Ikner's trial is scheduled for October 19, 2026. The investigation into OpenAI could develop in three directions. First: Florida fails — the criminal investigation runs into incompatibility with the traditional concept of culpability, and the case closes without a legal precedent. Second: The court recognizes a new category of liability for AI systems that actively generate content with harmful potential — which would fundamentally change the regulatory landscape for the entire industry. Third: OpenAI settles out of court, the case disappears without a public verdict, but political pressure for legal restrictions on AI chatbots increases significantly.