Loading page…
The planned California action challenges OpenAI’s decision not to contact police about a flagged account—a decision the company says its revised protocol would now handle differently.
Listen to this story
British Columbia plans to sue OpenAI in California over whether the company should have alerted authorities before the February 10 Tumbler Ridge Secondary School attack, which killed eight and wounded 27. OpenAI says it detected a policy violation in June 2025, closed the account after human review, and lacked evidence of credible, imminent planning.
OpenAI says its current referral protocol would escalate the same activity today.
British Columbia alleges safety teams flagged violent prompts but failed to contact the RCMP or local authorities.
The province’s planned case is separate from civil claims already filed by victims in California.
British Columbia says it will sue OpenAI in California, alleging the company failed to alert law enforcement to threats made through ChatGPT before the February attack at Tumbler Ridge Secondary School. The planned government action puts OpenAI’s internal threshold for referring violent activity to police at the center of the dispute.
Attorney General Niki Sharma announced the planned action on September 21. According to the province, the February 10 attack killed eight people and wounded 27. British Columbia alleges that OpenAI safety teams flagged violent prompts from the alleged perpetrator but did not notify the Royal Canadian Mounted Police or local authorities.
OpenAI says automated systems detected a usage-policy violation in June 2025, after which human reviewers shut down the account. The company says the information available then did not show credible, imminent planning that met its threshold for a law-enforcement referral. That assessment is now directly at odds with the province’s allegation that authorities should have been alerted.
OpenAI says that, under its enhanced law-enforcement referral protocol, it would refer the account to law enforcement if the same activity were discovered today. The company has also committed to build direct contact points with Canadian law enforcement, improve detection of repeat policy violators, expand localized crisis-support referrals and reassess its violence-detection thresholds.
Victims had already brought multiple civil claims against OpenAI in California. British Columbia says its planned action is separate from those cases and seeks accountability as well as support for rebuilding the community, including construction of a new Tumbler Ridge school facility.
The immediate issue is not whether OpenAI had a policy for violent activity, but whether closing an account without notifying authorities was adequate in this instance. Any future court action could test how much legal weight a company’s own referral threshold carries when government authorities argue a warning was warranted.
Editorial analysis
This is a meaningful escalation from the earlier private litigation: a provincial government is now pursuing its own accountability effort over how an AI company drew the line for contacting police. OpenAI’s statement that it would refer the account under its current protocol sharpens the dispute without resolving whether its earlier judgment created legal responsibility. The next important development is whether British Columbia’s planned action defines a concrete duty for platforms that detect violent activity but do not consider it imminent enough for referral.
Loading discussion...
Join the conversation
Explain where you would draw the reporting line.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.