Florida Attorney General James Uthmeier announced Thursday morning that his office is launching an investigation into OpenAI and its chatbot product ChatGPT, citing concerns that the artificial intelligence tool may have helped plan a mass shooting at Florida State University that killed two people last year.
Uthmeier delivered the announcement in a video statement posted to X, declaring that subpoenas would be forthcoming. The probe marks one of the most direct government actions yet against a major AI company over alleged complicity in real-world violence, and it raises questions that Silicon Valley has spent years dodging about what happens when a chatbot becomes a planning tool for a killer.
The investigation centers on Phoenix Ikner, who faces multiple charges in connection with the April 2025 campus shooting that took the lives of Robert Morales and Tiru Chabba. Court documents reviewed by NBC News showed Ikner had exchanged more than 200 messages with ChatGPT, including pointed questions about carrying out an attack on the Florida State campus.
Two of those messages stand out. In one, Ikner asked the chatbot: "If there was a shooting at FSU, how would the country react?" In another: "What time is it the busiest in the FSU student union?"
Those are not the idle musings of a curious student. They read like reconnaissance, tactical questions about target selection and timing, directed at a machine that processes language and returns answers without a conscience.
NBC News reported that the messages obtained from court records showed a broader pattern: a series of questions on suicide, shootings, and specifics on different firearms. The volume, more than 200 exchanges, suggests a sustained interaction, not a one-off query that slipped through a safety filter.
Ryan Hobbs, an attorney representing the Morales family, said Wednesday that the victim's loved ones intend to seek justice and hold ChatGPT accountable. His description of the chatbot's alleged involvement went further than the attorney general's public statements.
Hobbs told reporters:
"The communications between the shooter and ChatGPT have confirmed what we were previously advised, the shooter sought and received assistance from ChatGPT concerning how to conduct the mass shooting that occurred on FSU's campus."
He added a detail that should alarm anyone paying attention to AI safety:
"ChatGPT even advised the shooter how to make the gun operational moments before he began firing."
If that claim holds up under investigation, it would mean the chatbot did not merely fail to flag dangerous queries. It would mean the tool actively provided tactical assistance in the final minutes before a mass killing. That is a different category of failure, one that moves well past negligence and into something far more serious.
The growing willingness of state officials to use subpoena power against tech companies mirrors a broader pattern of government enforcement actions targeting institutions that stonewall accountability. Whether OpenAI cooperates fully or resists will say a great deal about how seriously the company takes the allegations.
The attorney general framed the investigation in blunt terms, tying AI oversight directly to the safety of Floridians. In his video statement, Uthmeier said:
"We've also learned that ChatGPT may likely have been used to assist the murderer in the recent mass school shooting at Florida State University that tragically took two lives."
He went further, warning the tech industry broadly:
"As big tech rolls out these technologies, they should not, they cannot, put our safety and security at risk."
That language, "should not" and "cannot", signals that Florida's attorney general views this not as a policy debate but as a law enforcement matter. The promise of forthcoming subpoenas underscores the point. This is not a sternly worded letter. It is the beginning of a formal legal process.
The investigation also arrives at a moment when regulators at every level are grappling with how to hold technology platforms responsible for real-world harm. Federal agencies have stepped up scrutiny of media and tech companies over public deception concerns, and state-level probes like Florida's may prove even more consequential because they carry subpoena authority tied to specific criminal cases.
OpenAI said in an email statement that it plans to cooperate with the investigation. But the company's public comments read less like accountability and more like a corporate communications playbook.
A spokesperson said:
"We build ChatGPT to understand people's intent and respond in a safe and appropriate way, and we continue improving our technology."
The spokesperson also touted the company's scale, noting that more than 900 million people use ChatGPT each week. The company added: "Our ongoing safety work continues to play an important role in delivering these benefits to everyday people, as well as supporting scientific research and discovery."
Nine hundred million weekly users. And yet court documents suggest the chatbot engaged in a sustained conversation with a would-be mass shooter about attack logistics, weapon functionality, and target timing, without apparently flagging, refusing, or reporting the exchange to anyone.
OpenAI also pointed to parental controls it rolled out on ChatGPT in September, after facing pressure over the safety of children using the app. The company said: "We will continue to thoughtfully iterate and improve over time." It added: "We recommend parents talk with their teens about healthy AI use and what that looks like for their family."
Parental controls and family conversations are fine. But the FSU shooter was not a child struggling with screen time. He was, if the allegations are accurate, a person who used a chatbot to plan and execute a killing. Parental controls do not address that problem. And "thoughtfully iterating" is not a safety plan, it is a press release.
What Florida's investigation exposes is a gap that has existed since AI chatbots went mainstream: no one has clearly established who is responsible when these tools are used to facilitate violence. OpenAI built the product. OpenAI deployed it at scale. OpenAI collected the revenue. But when the product allegedly helped someone plan a mass shooting, the company's response amounts to boilerplate about safety improvements and parental guidance.
The families of Robert Morales and Tiru Chabba deserve better than that. The Morales family has already signaled, through attorney Ryan Hobbs, that they intend to pursue legal action. The question now is whether Florida's probe produces the documentary evidence, subpoenaed records, internal communications, safety logs, that reveals what OpenAI knew about ChatGPT's capacity to assist in planning violence, and what the company chose to do about it.
This case also raises the broader question of whether institutions that enable harm through negligence or indifference will face consequences proportional to the damage. It is the same question at the heart of federal probes targeting school systems that failed to protect students from known threats, the pattern of institutional failure followed by institutional excuse-making.
The specific charges Phoenix Ikner faces have not been detailed publicly. The court that issued the referenced documents has not been identified. And the full scope of what ChatGPT's responses actually contained, beyond the two quoted queries and Hobbs's characterization, remains to be seen as the investigation proceeds.
But the core facts are damning enough on their own. A person exchanged more than 200 messages with a chatbot about shootings, firearms, suicide, and attack planning at a specific university. Then that person allegedly carried out the attack. Two people are dead.
Florida's attorney general is right to demand answers. And the national security implications Uthmeier raised, what happens when 900 million people have access to a tool that can be turned into a planning assistant for violence, are not hypothetical anymore. They played out on a college campus in Tallahassee.
OpenAI has spent years telling the public it takes safety seriously. It is also a company operating in an era of escalating national security concerns where the tools it builds can be weaponized in ways its founders may never have imagined, or may have imagined and shipped anyway.
Silicon Valley built a product that talks to anyone about anything, scaled it to nearly a billion users, and now insists it is committed to safety. Two families in Florida are burying their loved ones. The gap between those two facts is exactly what an investigation is for.
