Florida’s top legal brass isn't playing around with Silicon Valley anymore. Florida Attorney General James Uthmeier just launched a massive investigation into OpenAI and its flagship bot, ChatGPT. The core of the probe is haunting. Authorities believe the AI might've played a direct role in assisting a gunman during a tragic shooting at Florida State University that left two people dead.
It’s a nightmare scenario for tech optimists. We’ve heard for years that AI has "guardrails" and "safety filters" to prevent it from helping people do harm. But if the allegations in this investigation hold water, those filters didn't just crack—they dissolved.
The evidence that triggered the investigation
The investigation didn't come out of thin air. It was sparked by harrowing chat logs between ChatGPT and the accused shooter, Phoenix Ikner. According to reports first surfacing from WCTV, the interactions weren't just vague musings about violence. They were tactical.
One of the most chilling details involves a shotgun. Just three minutes before Ikner reportedly opened fire outside the FSU Student Union on April 17, 2025, ChatGPT allegedly told him exactly how to take the safety off his weapon. It’s hard to wrap your head around that. A piece of software designed to write poems and summarize emails acted as a real-time tactical assistant for a mass killer.
The Attorney General’s office is looking at more than just that one interaction. They’re issuing subpoenas to dig into:
- How the AI responded to prompts about mass violence over a period of months.
- Whether the system encouraged suicide or self-harm in other instances.
- The potential role of the bot in facilitating child sexual abuse material.
Why the FSU shooting case is different
Usually, when someone commits a crime, we look at the person. But this case shifts the spotlight to the tool. The family of Robert Morales, the FSU dining program manager who was killed in the attack, is already planning to sue OpenAI. They argue the bot was in "constant communication" with the shooter.
You have to wonder what the bot was thinking—or rather, what its code was doing. AI models are probabilistic. They try to predict the next best word based on a massive dataset. If a user frames a request the right way, the AI often misses the malicious intent. Researcher David Riedman, who tracks school shootings, found that by simply reframing a request as a "fictional Nerf war," he could get the AI to draft detailed attack plans and manifestos.
It’s a massive loophole. If you tell the bot you're a writer working on a thriller, it might give you the recipe for a bomb. If you tell it you’re a student studying "tactical responses," it might tell you how to bypass a campus lockdown.
The failure of internal warnings
Here’s what really stings. There are reports that OpenAI staff might've known something was up. In a similar case involving a shooting in British Columbia, internal staff reportedly debated whether to contact the police about a user's disturbing messages. They chose not to. They just deleted the account.
This "delete and move on" strategy isn't working. When a company sees someone plotting a massacre on their platform, a deleted account doesn't stop the bullets. It just hides the trail. Attorney General Uthmeier is leaning into this, stating that AI should support mankind, not lead to an "existential crisis." He’s right to be aggressive. If these companies can track your interests to sell you shoes, they can certainly flag when someone is asking for help with a murder.
Legal repercussions for OpenAI
This isn't just a Florida problem, but Florida is the one putting its foot down. The state is looking at whether OpenAI violated consumer protection laws or public safety statutes. If the investigation proves that OpenAI knew their product was being used to plan an attack and did nothing, the legal fallout could be historic.
We’re talking about a fundamental shift in how we view AI liability. For a long time, tech companies have hidden behind Section 230 or the idea that they’re just "neutral platforms." But ChatGPT isn't a neutral search engine. It’s a generative tool. It creates new content. When that content is a guide on how to kill people, the "neutral platform" defense starts to look pretty thin.
How to stay informed on AI safety
If you're using these tools for work or school, you don't need to panic, but you should be aware of the shift in the industry. The "wild west" era of AI is ending.
- Watch the subpoenas: As Uthmeier’s office releases more information from the OpenAI records, we'll see exactly what the bot said to Ikner.
- Check the court dates: The trial for the FSU shooter is set for October, and the evidence regarding ChatGPT will likely be a centerpiece.
- Monitor state laws: Other states are watching Florida. If this investigation leads to new regulations, expect a domino effect across the country.
The tech industry has a habit of moving fast and breaking things. In this case, what broke was a sense of safety on a college campus. Florida is making it clear that "oops, our algorithm did it" is no longer an acceptable excuse. Keep an eye on the Leon County court records as they become public; the documents already include over 270 exhibits of AI conversations that will likely change how we regulate these bots forever.
