A new investigation out of Florida is raising a question that would have sounded almost impossible not long ago: can an AI company face criminal scrutiny over what its chatbot said before a mass shooting?
That is the issue at the center of a report from Scott Budman of NBC Bay Area, who said Florida’s attorney general has now launched a criminal probe into OpenAI, the San Francisco-based company behind ChatGPT, over whether the chatbot may have helped a gunman plan part of the deadly shooting at Florida State University.
Budman described it as the kind of case the tech world has been watching for, even if few expected it to arrive this quickly. The reason is obvious. If prosecutors seriously try to link an AI system’s answers to a real-world killing, it could open a legal fight far bigger than one company or one case.
And that is why this story has the full attention of the tech industry. It is not only about one awful crime. It is about where responsibility begins and ends when a machine gives answers that a human later uses in the worst possible way.
The Case That Pushed This Fight Into Public View
Budman told viewers that the background starts with a 21-year-old suspect arrested in connection with a shooting at Florida State a year ago that left two people dead.
According to his report, investigators found that the suspect had been going back and forth with ChatGPT shortly before the shooting. Budman said the exchange included disturbing questions, including one about what time of day the most people would be in the student union.
That detail is what gives the case so much force. It is one thing for a shooter to search the internet in some broad way before an attack. It is another for prosecutors to say there was a direct conversation with a chatbot in the minutes leading up to violence.
Budman said that looked “very damaging,” and it is easy to see why. The allegation does not sound abstract. It sounds targeted, specific, and close in time to the killings themselves.
Florida Attorney General James Uthmeier has now taken the next step, with Budman reporting that subpoenas are being sent to OpenAI’s headquarters in San Francisco. The goal, he said, is to determine whether the chatbot gave ideas, information, or some form of planning assistance to the suspect who later killed people on campus.
Why Budman Says This Could Be Groundbreaking
NBC Bay Area anchor Janelle Wang called it a potentially groundbreaking case, and Budman agreed.
His point was that while people have heard for years about internet searches tied to crimes, this is different because it puts an interactive AI system under a criminal microscope. Search engines and video platforms have long faced criticism when users look up dangerous material, but Budman said tech companies have generally been protected from direct liability when the crime is ultimately committed by a human being.
That legal separation has mattered a lot. The usual argument is that the company provided access to information, while the individual made the criminal choice.
What Florida appears to be testing now is whether that line still holds when the system is not just passively hosting information, but actively responding to a user’s questions in a conversational format.
That is what makes this case feel new. A chatbot is not exactly a search bar, and it is not exactly a person either. It occupies a strange middle ground, and the law has not fully settled what to do with that.
In that sense, Budman is probably right to call it groundbreaking. Even if the case goes nowhere, the question it raises is not going away.
OpenAI’s Defense Is Already Taking Shape
Budman said OpenAI responded quickly once the issue surfaced publicly.
According to the company’s statement to NBC, OpenAI called the Florida State shooting a tragedy and said it felt deeply bad about the case. But the company also insisted that neither OpenAI nor ChatGPT broke the law.
As Budman explained it, OpenAI’s defense is built around a familiar argument: the information the suspect received was publicly available. In other words, the company appears to be saying that the chatbot did not create some hidden or exclusive guidance. It gave answers that could have been found through ordinary online sources.
That is a serious defense, and it is likely to matter a lot. If the information exists all over the internet already, OpenAI will argue that the blame rests entirely with the person who used it for evil.
Still, this is where the case gets morally uncomfortable, even before it becomes legally difficult. A company may be right that the facts were already public, but people will still ask whether a chatbot should have responded at all, and whether a more careful system would have recognized what kind of conversation it was entering.
That gap between what is technically allowed and what feels obviously dangerous is where a lot of the AI debate now lives.
What Florida Says It Is Looking For
Budman’s report made clear that Florida is not only asking what the chatbot said. Investigators also want to know how OpenAI handles threats more broadly.
The subpoenas, according to the material behind the NBC report, are expected to seek information about OpenAI’s policies and internal training materials dealing with user threats of harm to themselves or others. The state also wants to examine how the company works with law enforcement and how it reports suspected criminal behavior.
That suggests the probe is bigger than one conversation.
Florida appears to be asking whether OpenAI had systems in place to catch warning signs, whether those systems failed, and whether the company’s people knew enough to intervene or at least alert authorities. Budman summed it up as an effort to find out whether the company can be made culpable.
Attorney General Uthmeier put the argument in especially blunt terms when he said that if this had been a person on the other side of the screen, prosecutors would be looking at murder charges. That is a dramatic statement, and it will likely become one of the defining lines in this case.
It also shows how aggressive Florida wants to be. The state is not treating this as a vague ethical concern. It is openly testing whether criminal accountability can reach beyond the shooter himself.
The Ethics Question May Outrun The Legal One
Budman said he also spoke with a legal and ethics expert, and her view added another layer.
As he explained it, the ethical side of this case is fairly clear. It is wrong for someone to go online looking for information to do something violent. It is also obvious that AI companies do not want to be in a position where their products can be described as having helped someone kill people.
But the expert also told Budman that these companies could build stronger guardrails.
That may be the most practical point in the whole report. Even if OpenAI ultimately wins the legal fight, this case may still pressure the entire AI industry to tighten its systems. If a chatbot can be prompted in ways that meaningfully assist someone moving toward violence, companies will face rising pressure to catch those patterns earlier and shut them down more aggressively.
Budman said she expects that to happen in the future, possibly because of this very case.
That feels likely, whether through lawsuits, legislation, or simple corporate fear. Once a tragedy becomes tied this closely to a major AI product, the pressure to show stronger safeguards rises almost overnight.
The Bigger Problem This Case Exposes
What makes Budman’s report so important is that it does not present this as a one-off tech scandal. It presents it as a sign of where the next major fight is headed.
For years, the debate over online harm focused on social media, search engines, and video platforms. Now AI is moving into that same space, but with a more intimate role. These systems do not just display links. They answer, refine, suggest, and respond in real time.
That makes the stakes higher.
It also makes the legal questions harder. If ChatGPT gave factual answers drawn from public information, OpenAI will say it cannot be blamed for what a killer did next. Florida, on the other hand, seems ready to ask whether the company’s tools crossed from information into assistance.
That line may define this entire case.
For now, as Scott Budman reported, the subpoenas are going out and the criminal probe is underway. OpenAI says it did not break the law. Florida says the communications were serious enough to justify a criminal investigation.
And somewhere in the middle of that fight is a question that may shape the next era of tech law: when an AI system gives dangerous answers to a dangerous person, who, if anyone, is supposed to answer for it?
Raised in a small Arizona town, Kevin grew up surrounded by rugged desert landscapes and a family of hunters. His background in competitive shooting and firearms training has made him an authority on self-defense and gun safety. A certified firearms instructor, Kevin teaches others how to properly handle and maintain their weapons, whether for hunting, home defense, or survival situations. His writing focuses on responsible gun ownership, marksmanship, and the role of firearms in personal preparedness.
