Rushlow v. Altman
A grade 7 teacher locked her classroom door and hid her students underneath her desk while a gunman fired through it. She is suing OpenAI and Sam Altman, saying the company was told eight months earlier that he was dangerous and decided not to call the police.
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, no court has ruled on any of it, and the attacker died at the scene and no criminal case followed.
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.
- February 10, 2026. An 18-year-old killed his mother and 11-year-old brother at home, then drove to Tumbler Ridge Secondary School in British Columbia and opened fire. Five children and an education assistant were killed. Two students were severely wounded. The attacker died at the scene.
- June 2025, eight months earlier. The complaint says an automated OpenAI system flagged his ChatGPT account over what he was writing about guns, and that trained reviewers concluded he was a real threat and recommended calling the Royal Canadian Mounted Police.
- The decision. The complaint says company leadership overruled those reviewers, shut the account down, and never made the call. It says he opened another account and kept using the product.
- April 24, 2026. Sam Altman published a letter saying he was deeply sorry the company did not alert law enforcement to the account.
- September 2, 2026. Deidre Rushlow filed this case in San Francisco federal court.
The story.
Deidre Rushlow taught grade 7. Earlier that period she had sent part of her class to the library with an education assistant. Minutes later the attacker reached that library.
When she heard gunfire and screaming down the hall, she locked her door and hid the students who were still with her under her desk. One of them was her nephew. She phoned the library and texted the assistant to find out whether the children she had just sent there were safe. Nobody answered.
The attacker reached her room, found it locked, and fired through the front door, then walked around and fired through the back. The complaint says the rounds struck the wall behind where she and her students were hiding. She texted her husband a goodbye.
She survived. The complaint is about what that did to her, and about a decision made eight months earlier in San Francisco.
What the lawsuit says.
The core allegation is that this was preventable and that OpenAI knew it.
According to the complaint, the account was routed to a specialist team inside OpenAI whose job is to review users who appear to be planning to harm people. Those reviewers decided he was a credible threat. More than one of them recommended contacting the RCMP.
The complaint says that recommendation went up the chain and was rejected, on the basis that the case did not meet the company's threshold for a credible and imminent risk. It also says the people who made that call had no training in threat assessment, while the people they overruled did.
It adds that the RCMP already had a file on him. Officers had been to the house more than once over mental health concerns and had temporarily removed firearms. A referral would have reached people who already knew who he was.
Deactivating an account is not the same as banning someone.
After the shooting, OpenAI said it had banned the account. The complaint says the company has no way to ban a person at all. What it has is deactivation, which closes one email address.
It points to OpenAI's own help documentation, which tells deactivated users they can come back by registering with a different address, or by adding a suffix to the one they already have. On that account, describing his return as evading safeguards describes a route the company published itself.
The comparison the complaint keeps coming back to.
This is the part that does not depend on reading anyone's chat logs.
The complaint says threats to OpenAI's own staff are handled by a separate corporate security team that reports to the chief operating officer and can call the police on its own authority. It points to November 21, 2025, when the company locked down its offices, warned employees, circulated a suspect's name and photograph, and called the San Francisco Police Department, while saying there was no sign of any active threat.
Set against the refusal to call the RCMP about a flagged user, the complaint uses that contrast to argue the company's standards bend depending on who is at risk.
OpenAI disputes the plaintiffs' account of how these decisions are made.
What she is asking for.
The complaint pleads nine counts. Behind the labels they come down to a few arguments.
- That OpenAI was careless in three separate ways: in not warning police, in not warning the public about how the product behaves, and in how it built the product.
- That the product itself was defective, both in design and because it was sold without adequate warnings.
- That handing the product back to someone the company already knew was dangerous was itself wrong.
- That once OpenAI chose to review the account, it had to do that job carefully, and did not.
- That she should be compensated for the emotional harm of being in the path of the attack.
She is also asking the court to order changes at OpenAI, including real bans for users removed over violence, review of new accounts linked to them, notifying police when internal systems identify someone who poses a real-world risk, and independent monitoring.
Who this may affect.
Most of the cases against OpenAI have been brought by people harmed by their own use of the product, or by their families. This one was brought by someone who never used it.
If it succeeds, it widens who can bring a claim to include people caught up in harm caused by somebody else's use. That is a much larger group, and it is the reason this case is worth watching even if the details do not match your situation.
It also turns on what a company owes strangers it has never dealt with. OpenAI will argue it owed nothing to a town in British Columbia it had never heard of. The complaint's answer is that the company chose to look at this user, and that choosing to look created a responsibility to act.
Common questions.
What is this lawsuit about?
Deidre Rushlow is a grade 7 teacher who was inside Tumbler Ridge Secondary School during the February 10, 2026 mass shooting. She sued Sam Altman and three OpenAI companies in San Francisco federal court on September 2, 2026, saying OpenAI's own staff flagged the attacker's ChatGPT account eight months earlier and recommended calling the police, and that company leadership overruled them.
Can I join this lawsuit?
No. This is one person's case, not a class action, so there is nothing to join and no claims process. Whether your own situation supports a claim depends on your facts, your location, and your deadline. A licensed attorney has to answer that.
Is there a settlement or payout?
No. The case was filed on September 2, 2026 and OpenAI has not responded. No court has ruled on anything. There is no settlement, no fund, and no compensation program tied to this case.
Why is a Canadian case in a United States court?
OpenAI is based in San Francisco. When someone from another country sues an American company, federal court is open to them, and the complaint relies on that. Suing the company where it is based is also the most direct route to its internal records.
How is this different from the earlier Tumbler Ridge cases?
Seven cases were filed in April 2026 by families of those killed and injured. This one is part of a second wave brought by people who were inside the building and survived. Her claim is for the harm of being in the path of the attack rather than for the loss of a family member.
What did Sam Altman say about it?
In an April 24, 2026 letter to the Tumbler Ridge community, he said he was deeply sorry the company did not alert law enforcement to the account. The complaint treats that letter as an admission. OpenAI disputes the rest of the account, including who made the decision and why.
How long do I have to file a claim?
Deadlines vary by location and are often shorter than people expect. Some start when you were hurt, others when you found out what caused it. Because the deadline can quietly run out, this is worth asking sooner rather than later.
Where can I read the actual filing?
The full complaint is posted above as a PDF, straight from the court file. It runs 48 pages. The case number is 3:26-cv-09344-PHK in the U.S. District Court for the Northern District of California, San Francisco Division.
Request a case review.
If you were injured in an incident where an AI chatbot may have played a role, whether you used it yourself or not, you can request a free, confidential case review. Reviews are handled by participating legal professionals and intake partners. There is no cost and no obligation, and submitting a request does not create an attorney-client relationship.
Continue exploring this category.
Gourley v. OpenAI
A survivor of the Florida State University shooting, with that complaint posted in full.
Winters v. OpenAI
The ChatGPT medical-advice case, with that complaint posted in full as well.
OpenAI Case Review
Case review specific to claims involving OpenAI and ChatGPT.
AI Chatbot Case Review
Case review options for harms involving AI chatbots and companion apps.
The Tumbler Ridge Cases, Explained
Attorney analysis of the wider set of cases arising from the same shooting.
Every OpenAI Case, Tracked
Status and background on all the pending OpenAI lawsuits, on Lawsuit Informer.
Browse All Lawsuits
Every litigation category tracked 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.