British Columbia sues OpenAI over ChatGPT-linked shooting

British Columbia and a local school board have sued OpenAI and CEO Sam Altman, demanding they fund a new school and emergency services following a mass shooting. On February 10, 2026, 18-year-old Jesse Van Rootselaar killed five children, an education assistant, their mother, and their half-brother before dying by suicide. The tragedy devastated Tumbler Ridge, a remote mining town of 2,700 people. The local secondary school was demolished in August due to severe trauma, forcing 160 students and staff into temporary facilities initially built from 14 Ministry of Forestry trailers.

The lawsuit alleges that OpenAI's negligence directly contributed to the tragedy. Whistleblowers revealed that OpenAI's human reviewers flagged the shooter's violent ChatGPT conversations in June 2025, identifying a credible threat of gun violence. However, company leadership allegedly overruled safety teams and deactivated the account instead of notifying the Royal Canadian Mounted Police, allowing the shooter to create a second account. The province claims OpenAI prioritized protecting its potential $1 trillion IPO valuation over public safety, ignoring recurring patterns of violence previously linked to ChatGPT, including a Florida State University shooting and a Las Vegas bombing.

In addition to funding the new school and covering extensive healthcare costs—including over 900 counseling sessions—the plaintiffs want the court to force OpenAI to release the shooter's chat logs. They also demand injunctive relief to alter how ChatGPT operates. Specifically, the lawsuit targets core instructions directing the system to assume "good faith" and "not to probe intent" during violent conversations. The province wants these rules changed so the AI automatically terminates violent chats and triggers law enforcement alerts.

For AI practitioners, this litigation highlights the escalating legal liabilities associated with safety alignment. If the court rules against OpenAI, it could establish a precedent that AI developers hold a duty of care to report imminent threats. Developers may need to redesign system prompts, move away from sycophantic model behaviors, and implement robust, independent safety audits to avoid massive punitive damages and liability for real-world violence.