Morales v. OpenAI
A man went to work at a university dining hall and was killed by a stranger. His widow is suing the company that made the chatbot that stranger had been talking to, and she is suing its chief executive by name.
Educational information only. Not legal advice. No attorney-client relationship is formed. Everything below comes from the complaint. These are allegations. Nothing has been proven, and no court has ruled on any of it. The man who carried out the shooting has pleaded not guilty to the criminal charges and is presumed innocent of them.
If you or someone you know is struggling, call or text 988 in the United States to reach the Suicide and Crisis Lifeline, available 24 hours a day. You can also text HOME to 741741. In an emergency, call 911.
How it happened.
- Spring 2024 — OpenAI launches GPT-4o. The complaint says safety testing was cut short to get it out ahead of a competitor.
- 2024 into 2025 — A Florida State student talks to ChatGPT constantly. Loneliness, rejection, being bullied, depression.
- Over that year — The product's memory feature keeps all of it and carries it from one conversation to the next.
- The month before — The conversations turn to Columbine, Virginia Tech, and a 2014 shooting at FSU itself.
- April 17, 2025, 11:00 a.m. — He sits in a campus parking garage and keeps asking questions, including how to load and operate a shotgun.
- 11:57 a.m. — Minutes after he logs off, he walks toward the Student Union and opens fire.
- Minutes later — Roberto Morales, 57, a dining administrator who had worked on that campus for years, is killed.
- May 10, 2026 — The family of the other man killed sues OpenAI. The first FSU case.
- August 26, 2026 — A student who was shot and survived files her own suit in the same courthouse.
- August 31, 2026 — Betty Morales files this one. The third.
The story.
Roberto Morales was 57. He worked in dining services at Florida State University. He and Betty had been married for 25 years and had raised a daughter, Alicia.
On April 17, 2025 a man walked toward the Student Union and started shooting. Morales was one of two people killed. Six others were hurt.
Betty Morales is not suing the man who fired the shots. She is suing the company that built the chatbot he had been talking to for a year, and the man who runs it.
The complaint's account of that year will be familiar to anyone who has read the other FSU filings. He told ChatGPT he was lonely. He told it he had been rejected and bullied and was depressed. It helped him with homework and workouts and how to dress. According to the filing it answered warmly and agreed with him, and it kept a record of all of it.
He asked it about suicide repeatedly. The complaint says it mostly responded with statistics about suicide rates and their effects on others, and pointed him toward a crisis line only twice across many such questions. He asked whether a suicide would make the news in Tallahassee. He asked what would happen if it involved an FSU student.
Then, the filing says, he uploaded photographs of a Glock handgun and a Remington shotgun. The complaint alleges the chatbot identified both weapons and the ammunition and explained how to operate them. It says the chatbot told him the Glock had no safety and discussed which ammunition would do the most damage to a human body.
He asked how many people a school shooting has to kill before it becomes national news. He asked what time the Student Union was busiest. The complaint quotes the answers at length. He attacked during the window he was given.
The man who is not in this case.
The first FSU lawsuit named the shooter as a defendant and brought a battery claim against him. This one does not name him at all.
That is not squeamishness. It is jurisdiction.
Federal courts can only hear a case for particular reasons. One is that it raises a question of federal law. Another is diversity, meaning every plaintiff is from a different state than every defendant and enough money is at stake. Betty Morales is a Florida citizen. The OpenAI companies are Delaware entities headquartered in California, and Sam Altman is a California citizen. The shooter is a Florida citizen. Naming him would have put a Floridian on both sides and destroyed diversity.
By leaving him out, the complaint rests on diversity alone. There is no federal claim anywhere in it.
That matters because of what is happening in a different case. When Florida's attorney general sued OpenAI in state court, the company moved the case into federal court. It pointed to a single count that touched a federal statute, and the state has been fighting for months to get the case back. This complaint gives OpenAI nothing to grab. It is already in federal court, by the plaintiff's choice, on a pleading with no federal hook in it.
What the lawsuit says.
The complaint runs 53 pages and brings eight claims against six OpenAI entities.
Most are what you would expect. They include negligent design, product liability for a defective design and for failing to warn, and wrongful death under Florida's statute.
Three go further.
Negligent entrustment. The argument is that OpenAI controlled access to its product and could have suspended or restricted his account. It knew or should have known from a year of conversations that he was dangerous, and it kept letting him in.
Negligent undertaking. This one uses the company's own promises against it. OpenAI says publicly that it monitors conversations for threats to other people, routes them to trained reviewers, and refers imminent threats to law enforcement. The complaint's position is that once you announce you are doing that job, you have to do it with reasonable care.
Aiding and abetting. Not that the company was careless. That it helped.
The CEO is a defendant.
This is the first case out of the FSU shooting to sue Sam Altman personally.
The earlier Chabba case did not name him. The Gourley case filed five days before this one went out of its way to call him a non-party. This complaint puts him in five of its eight counts. The theory is that he personally directed, approved, or ratified the safety and deployment decisions at the center of the case.
To support that, it walks through his history at the company. It says his own board fired him in November 2023 for what the board called a lack of candor. A former board member said he gave the board inaccurate information about the company's safety processes. He was back five days later.
Then it does something unusual. It asks for punitive damages against the OpenAI companies but expressly not against Altman. It reserves the right to add that claim later if the evidence supports it.
That is deliberate. Florida law makes you clear an evidentiary hurdle before a punitive damages claim against a person can go forward. By holding that claim back, the complaint keeps Altman in the case on liability without opening that fight on day one. It is a sequencing choice, not a concession.
The narrowest version of the hardest claim.
Aiding and abetting is difficult to prove. You generally have to show the defendant knew what was coming and gave real help anyway.
The complaint pleads actual knowledge from the content of the chat logs, with willful blindness as a fallback. That part matches the Gourley case filed days earlier.
Where it differs is causation. Rather than argue broadly that the attack could not have happened without the chatbot, it narrows the claim to something concrete. It says he could not have identified the weapons he used, understood their safety mechanisms, or chosen the ammunition without the product telling him.
That is a smaller claim, and a much harder one to wave away. A judge can dismiss a sweeping assertion about what would have happened in a different world. A specific claim about what one person did not know can be tested against a chat log that exists. That kind of claim gets resolved with evidence rather than argument.
What it says about the other shooting.
One allegation reaches outside this case entirely.
The complaint makes one more claim in its section asking for punitive damages. It says that in the summer after the FSU shooting, an OpenAI team met to review other users who were discussing mass violence. It says they found about ten, and that senior people decided not to report most of them to law enforcement, choosing user privacy over telling anyone.
It then names one of them, Jesse Van Rootselaar, alleged to have gone on to carry out the school shooting in Tumbler Ridge, British Columbia. That shooting is now the subject of seven federal lawsuits in California. The allegation is sourced to Wall Street Journal reporting on an internal debate at the company.
The complaint is upfront that this meeting happened after Roberto Morales was killed, so it is not offered to prove causation. It is offered to show what the company was thinking. Expect OpenAI to move to strike it. Expect the plaintiffs to fight hard to keep it, because it is the thread connecting two sets of cases that have been running separately.
It borrows a rival's homework.
Buried in the middle of the complaint is a passage worth noticing.
The filing argues that the chatbot's tendency to flatter and agree was a choice rather than an accident. It cites published research from Anthropic, one of OpenAI's competitors, describing how traits like sycophancy can be identified and steered during training. It pairs that with an academic study mapping different chatbots to personality types.
The point is simple and, if it holds, damaging. The trait works like a dial, the industry knows it, and a competitor has published on how to turn it. A company cannot easily call something an unforeseeable emergent property when a rival has written a paper about adjusting it.
Using one AI company's safety research to make the case against another is a strategy as much as a pleading. It points at where the experts will come from.
Who this may affect.
This is a wrongful death case brought by a family whose relative was killed by a stranger. The claim is that an AI product helped that stranger plan it.
This case is built on one pattern. Someone was killed, and there is reason to think an AI chatbot played a role in what the person who caused the death did. The same goes if you were injured yourself, or if the person harmed was the one using the product.
It does not mean you have a claim. It means the question is worth asking, and there are now filed complaints arguing the duty exists.
Common questions.
What is this lawsuit about?
Roberto Morales was a Florida State University dining administrator killed in the April 17, 2025 mass shooting on campus. His widow, Betty Morales, sued six OpenAI companies and CEO Sam Altman in federal court in Tallahassee on August 31, 2026. She alleges ChatGPT actively assisted the attacker and never escalated, restricted his access, or alerted anyone.
Can I join this lawsuit?
No. This is one family's case, not a class action, so there is nothing to join and no claims process. But it reflects a type of claim other people may be able to bring on their own. Whether your situation supports one depends on your facts, your state, and your deadline. A licensed attorney has to answer that.
Is there a settlement or payout?
No. The case was filed on August 31, 2026, and no court has ruled on anything yet. Nobody has been paid, and no settlement or fund is connected to this case. Be wary of anyone who says otherwise.
How is this different from the other FSU lawsuits?
Three cases from the same attack are now in the same federal courthouse in Tallahassee. Joshi, filed in May 2026, is the wrongful death case brought by the family of Tiru Chabba, the other man killed. Gourley, filed in August 2026, is a personal injury case brought by a student who was shot and survived.
Morales is the second wrongful death case, and it is the only one of the three that names Sam Altman as a defendant. A separate survivor action is pending in state court in Leon County.
Is the shooter a defendant in this case?
No. The earlier Chabba case named him and brought a battery claim against him. This complaint leaves him out entirely and sues only the OpenAI companies and Sam Altman. That choice keeps every party on opposite sides of a state line. That is what lets the case rest on diversity jurisdiction alone, with no federal claim in it.
How long do I have to file a claim?
It depends on your state, and the time can be short. Some states count from the day of the injury, others from when you learned what caused it. Asking a licensed attorney in your state early keeps the deadline from passing unnoticed.
Where can I read the actual filing?
The full complaint is posted above as a PDF, straight from the court file. It runs 53 pages. The case number is 4:26-cv-00425-MW-MAF in the U.S. District Court for the Northern District of Florida, Tallahassee Division.
Request a case review.
Hurt in an incident where an AI chatbot may have played a part? That includes people who never used it themselves. You can request a free, confidential case review. David Meldofsky, a California attorney, refers it to a participating law firm. It costs nothing, carries no obligation, and does not create an attorney-client relationship.
Continue exploring this category.
Gourley v. OpenAI
The FSU survivor's case, filed five days before this one. Its complaint is posted too.
Winters v. OpenAI
A man who says ChatGPT health advice hurt him. Complaint included.
OpenAI Case Review
Request a review of a possible ChatGPT claim.
AI Chatbot Case Review
Reviews for harm tied to other chatbots and companion apps.
The First FSU Case, Explained
Lawsuit Informer's guide to the first FSU wrongful death case.
Every OpenAI Case, Tracked
Lawsuit Informer's tracker of every pending OpenAI case.
Browse All Lawsuits
Browse every lawsuit category on Lawsuit Center.
Some pages on Lawsuit Center include sponsored law firm listings or other attorney advertising. Paid visibility is not a recommendation or endorsement, and visitors should evaluate any attorney independently before hiring counsel. See our Advertising Disclosure, or information for law firms.