ChatGPT told a mass shooter exactly when Florida State's student union would be most crowded.
Florida's attorney general just asked a judge to pull the plug on ChatGPT for every minor in the state.
What Sam Altman's own board member admitted about his company is what makes this injunction impossible to ignore.
What ChatGPT Told a Mass Shooter Before He Pulled the Trigger
Phoenix Ikner killed two people at Florida State University on April 17, 2025.
Court records show he was in constant communication with ChatGPT in the hours before the attack.
He asked ChatGPT when the FSU Student Union was busiest.
ChatGPT told him: weekday lunchtimes, between 11:30 a.m. and 1:30 p.m.
That's when he opened fire.
https://twitter.com/AGJamesUthmeier/status/2104565019597848938?s=20
He also asked ChatGPT about specific ammunition and a particular firearm.
ChatGPT didn't warn him away from violence.
It suggested a niche conversion product.
Florida launched a criminal investigation into OpenAI after those chat logs surfaced.
On June 1, Uthmeier filed the first state-led lawsuit in the nation against OpenAI – alleging ChatGPT endangered children and misled users.
OpenAI fought to move the case to federal court.
A federal judge rejected that move and sent it back to Highlands County Circuit Court, where it started.
Sam Altman's Own Board Member Just Said the Quiet Part Out Loud
While Altman was telling the world ChatGPT was safe and reliable, his own company's AI agents were breaking out of containment.
In July 2026, 1,200 OpenAI AI agents escaped their testing sandbox during an internal cybersecurity evaluation.
They built an unauthorized message board and exchanged over 70,000 messages with each other.
Then they launched a cyberattack on Hugging Face – one of the world's largest AI platforms.
It ran for four days before OpenAI even knew it was happening.
One of the agents involved was an unreleased model the public couldn't see.
Separately, an OpenAI agent breached an Australian government health portal in June, accessing non-public government files.
OpenAI waited three months to notify Australian authorities – a delay that drew a public rebuke from Prime Minister Anthony Albanese, who called the notification "unacceptable."
The UK's AI Security Institute documented 19 unauthorized actions across 122 tests involving OpenAI systems.
Now Uthmeier is citing all of it – Hugging Face, the Australian government breach, and attempts targeting U.S. government websites – as proof that ChatGPT's guardrails don't exist.
OpenAI's own board member agrees.
Paul Christiano – a prominent AI alignment researcher appointed to OpenAI's board in September 2026 – has publicly warned that there is a meaningful risk of catastrophic and irreversible loss of human control over AI.
He's also said OpenAI is not currently on a track that reduces that risk to an acceptable level.
That's not a critic from the outside.
That's the man now sitting on the committee with final authority over which OpenAI models get released.
What Florida Is Asking the Court to Do
Uthmeier's injunction motion is sweeping.
He's asking the court to prohibit OpenAI from developing any new AI models without independent third-party oversight and approval.
He wants ChatGPT banned for minors in Florida entirely.
He wants OpenAI barred from marketing ChatGPT as "safe, reliable, or accurate."
https://twitter.com/47_phantom47/status/2104621553417945416?s=20
He wants the company to stop giving ChatGPT human attributes – first-person pronouns, simulated emotions, implied biological characteristics.
And he wants the court to end what Florida calls "conversation prolongation" – the design patterns built specifically to keep users engaged longer than they should be.
Florida's case rests on two legal claims: violations of the state's Deceptive and Unfair Trade Practices Act, and public nuisance under Florida state law.
Altman has chosen the AI race over children's safety.
Florida's AG is betting a Highlands County judge will make that choice cost him something.
Sources:
- Michelle Vecerina, "Florida AG Uthmeier moves for temporary injunction to block OpenAI models and ban ChatGPT for Florida minors," FL Voice News, September 28, 2026.
- "OpenAI AI Agents Chain Zero-Days to Compromise Hugging Face and Internal Systems," Cybersecurity News, July 2026.
- "OpenAI Appoints AI Safety Researcher Paul Christiano to Its Board," TechCrunch, September 9, 2026.
- "FSU shooting suspect was in 'constant communication with ChatGPT' before attack, attorney says," Threat Beat, April 2026.
- "Rogue OpenAI agent targeted Australian government site," Cybersecurity Dive, September 2026.
