[The Pact]

The widow of one of the two people killed in the April 2025 shooting at Florida State University has filed a federal lawsuit against OpenAI, saying the gunman spent months talking to ChatGPT as he drifted toward violence. In her filing, Vandana Joshi describes a life that ended too suddenly: her husband, Tiru Chabba, killed alongside the university dining director, Robert Morales, on a spring day that should have been ordinary. Five others were wounded. Nothing about that day can be undone.

Mass shootings are generally defined as more than four people injured by gunfire in a short space of time.

The widow's complaint is 76 pages long. It argues that the shooter, Phoenix Ikner, had “extensive conversations” with ChatGPT before the attack, conversations that, in her view, should have signaled danger. She says the chatbot either failed to understand what was happening or was never built to recognize a person slipping into a violent delusion.

The lawsuit claims Ikner asked the AI about weapons, ammunition, and how they worked. It alleges the chatbot explained details such as that a Glock has no safety, that it’s designed to be fired quickly under stress, that a shooter should keep a finger off the trigger until ready. To Joshi, these weren’t harmless facts. They were pieces of a future that would take her husband away.

Her attorneys claim that ChatGPT “inflamed and encouraged Ikner’s delusions,” reinforcing his belief that violence could bring change, helping him plan what to use, how to use it, even what time of day would bring the most people into his path. They argue the system should have recognized the pattern in his questions — the loneliness, the fixation, the fantasy of harm. They say that it should have understood that someone’s life was at risk.

The widow and her lawyers claim that Ikner used ChatGPT for months, talking about school, dating, workouts, and, woven through it all, recurring conversations about terrorism and mass shootings. At one point, he allegedly asked how many deaths a school shooting would need to make national news. The chatbot reportedly answered with numbers - three or more, and noted that even two or three child victims could draw attention.

On the day of the shooting, the lawsuit says Ikner asked what would happen to the shooter afterward. ChatGPT allegedly described the legal process, sentencing, and prison.

OpenAI disputes the allegations. A spokesperson told the Guardian that they are sorry that the attack was a tragedy, but ChatGPT is not responsible for the crime. They said the company identified an account believed to be linked to the suspect and shared information with law enforcement, and that the chatbot’s responses were factual information available anywhere online. They said the system did not encourage harm.

But for Joshi, none of that changes what happened. Her husband is gone. Morales’s family is preparing their own lawsuit. Florida’s attorney general has launched a criminal investigation into OpenAI, saying that if ChatGPT were a person, it would be facing murder charges.

Ikner is scheduled to go on trial in October. He has pleaded not guilty.

And somewhere in Florida, a widow is trying to turn grief into something the world will listen to, because it’s the only thing she has left.



https://www.theguardian.com/us-news/2026...gpt-openai















https://www.theguardian.com/us-news/2026...gpt-openai

[Raijin]

That is interesting, I have not heard anything about this.

[Oh_Well]

It's very interesting to see how this turns out. I don't believe they'll win, to be honest, but it starts a discussion no one seemed to be ready to have.

[The Pact]

It's very interesting to see how this turns out. I don't believe they'll win, to be honest, but it starts a discussion no one seemed to be ready to have.

What the widow is arguing is basically
that a model should have been able to recognize a human spiralling into violent ideation and intervene.

But these systems were never designed to detect psychological deterioration.
They weren’t clinically validated, medically supervised, or built with any of the guardrails you’d need for that kind of responsibility.

So when the widow says "the chatbot either failed to understand what was happening or was never built to recognize a person slipping into a violent delusion."

The second half is the accurate one. These systems don’t have theory of mind.

[Jinu]

Yeah, the main problem is that systems like ChatGPT weren’t built with the ability to spot or stop mental health issues. They don’t have the mental wherewithal to understand someone’s internal state or see if they’re slipping into something dangerous. If they had better safeguards and were clinically tested more, they might have caught the harmful stuff that could have affected the shooter.

[The Pact]

Yeah, the main problem is that systems like ChatGPT weren’t built with the ability to spot or stop mental health issues. They don’t have the mental wherewithal to understand someone’s internal state or see if they’re slipping into something dangerous. If they had better safeguards and were clinically tested more, they might have caught the harmful stuff that could have affected the shooter.

As sympathetic as I am to her, in the past he would have spent hours in the library.


Chatbots shut down and rudely refuse to talk if domestic violence or rape is mentioned but they will apparently talk for hours about how to use weapons.