OpenAI’s own safety team flagged the account. Someone overruled a recommendation to call police. Months later, nine people were dead. Altman’s own written apology — “I am deeply sorry” — now anchors the province’s legal case against the company he leads.
This article discusses a school shooting and its aftermath at a legal and policy level.
British Columbia filed a civil lawsuit against OpenAI and CEO Sam Altman, in a US District Court in San Francisco, alleging the company had information that could have prevented the February 2026 shooting at Tumbler Ridge Secondary School but failed to act on it. Nine people died in the attack, including six children and an educational assistant, before the shooter, 18-year-old Jesse Van Rootselaar, died by suicide at the scene. BC Attorney General Niki Sharma, filing alongside the Peace River South School District, said Tumbler Ridge residents “deserve answers and accountability.”
What’s Happening & Why It Matters
What the Province Says OpenAI Knew, and When
The lawsuit’s central allegation is specific and dated. In June 2025, OpenAI’s own safety team flagged Van Rootselaar’s ChatGPT account after her conversations revealed repeated references to gun violence across multiple days. According to the province’s complaint, that flagging came with an internal recommendation to contact police — a recommendation that was overruled. OpenAI banned the account for the disturbing content it contained. It did not alert law enforcement. The attack happened eight months later, on 10 February 2026.
Sam Altman’s own written apology to the Tumbler Ridge community, sent in April 2026, now functions as a central piece of evidence in BC’s case. “I am deeply sorry that we did not alert law enforcement to the account that was banned in June,” Altman wrote. The province’s lawsuit characterises that statement as a direct admission: OpenAI identified a genuine risk, took action against the account itself, and still chose not to contact police.
Legal Theory: “No AI Exemption”
Sharma stated the province’s argument in narrow legal terms, rejecting any suggestion that AI companies operate under different rules than other institutions. “There is no AI exemption to established duties of care,” she said, grounding the case in existing negligence and product liability law rather than any AI-specific statute. That impacts how courts are likely to treat the case — BC isn’t asking judges to invent a new legal category for AI harm. It’s asking them to apply duty-of-care principles that already exist in wrongful-death and negligence law to a company that, by its own admission, identified a specific risk and didn’t escalate it.
Legal analysts have flagged the difficult evidentiary question sitting underneath. The province must demonstrate that a warning from OpenAI would have reached police in time to matter, and that the company owed a duty its internal reporting criteria displaced. Everything in the complaint about an overruled safety recommendation remains, for now, an allegation OpenAI has not formally admitted in court — Altman’s public apology addresses the failure to alert police, but doesn’t concede the specific internal decision-making process the lawsuit describes.
What British Columbia Is Asking For
The province’s lawsuit seeks two distinct forms of relief. First, financial compensation covering public costs tied to the response and recovery — demolition of the original school building, construction of its replacement, and the deployment of additional educators, clinicians, victim services workers, and police to support the community afterwards. Tumbler Ridge officials said they hope to build a replacement school at a new location by the end of 2026.
Second, and more consequential long-term, the province is seeking injunctive relief — a court order requiring OpenAI to change how it handles ChatGPT conversations that may indicate a risk of violence, including terminating conversations that cross that threshold. That second request extends well beyond compensating Tumbler Ridge. If granted, it would impose a court-mandated design change on how ChatGPT itself functions for every user, not just a financial settlement tied to this one case.
A Familiar Legal Track
BC’s provincial lawsuit runs parallel to, but separately from, a larger wave of family-led litigation already moving through the same court system. As of early September 2026, 37 civil suits had been filed against OpenAI and Altman by victims and families in the Northern District of California, with 30 additional plaintiffs joining that action earlier this month. OpenAI has declined to release Van Rootselaar’s full ChatGPT logs to the province. However, the company says it has provided the logs to law enforcement and offered to cooperate with plaintiffs during formal discovery.
This isn’t OpenAI’s active lawsuit alleging harm connected to ChatGPT conversations involving a minor. As documented in ongoing litigation, the company faces comparable wrongful-death claims tied to other user conversations that raised safety concerns OpenAI’s systems flagged but didn’t escalate to intervention. Florida became the first US state to sue OpenAI over safety risks in June 2026, accusing the company of misrepresenting ChatGPT’s safety — a separate but thematically related regulatory track BC’s own case now joins.
TF Summary: What’s Next
OpenAI has not filed a formal response to British Columbia’s complaint. The 37 family-led civil suits continue proceeding in the Northern District of California, alongside the filed provincial action. Discovery, including the full scope of Van Rootselaar’s ChatGPT conversation logs, remains a contested and unresolved issue between the province and OpenAI. No trial date has been set for either the provincial lawsuit or the family litigation.
MY FORECAST: Expect this case to become a genuine test of how courts apply existing product liability and negligence frameworks to AI companies — Sharma’s “no AI exemption” is a deliberate legal strategy designed to avoid the harder, novel question of whether AI needs its own liability category, and that strategic choice will shape how other jurisdictions structure comparable cases going forward. The discovery fight over Van Rootselaar’s full conversation logs will become the case’s most consequential near-term battle, given how the province’s entire theory depends on establishing what OpenAI knew, when, and what its internal escalation process required before it was overruled.
If you or someone you know needs support, please get in touch with the 988 Suicide and Crisis Lifeline by calling or texting 988 in the US, or the Canada Suicide Prevention Service at 1-833-456-4566.
Related Stories
- A Judge Rules Pentagon’s Anthropic Denylisting Illegal
- Meta Settles States’ Child Safety Trial for $17Bn
- OpenAI Discloses More ‘Concerning’ AI Behaviours