Months before the deadly mass shooting in Tumbler Ridge, British Columbia, OpenAI considered alerting authorities about Jesse Van Rootselaar’s activity on its ChatGPT platform. The Wall Street Journal reported that Van Rootselaar described gun violence scenarios over several days last June, triggering automated alerts.

OpenAI employees debated whether her posts indicated a real-world threat. Some urged notifying Canadian authorities, but company leaders ultimately decided against it, citing that the content did not meet the threshold of an imminent risk of serious harm. Van Rootselaar’s account was banned. Following the shooting, OpenAI contacted the RCMP and is assisting the investigation. “Our thoughts are with everyone affected by the Tumbler Ridge tragedy," the company said.

Van Rootselaar’s online footprint included a Roblox game simulating a mass shooting and social media posts on anime, drugs, and her gender transition. She was previously known to local police, who had visited her home multiple times for mental health concerns and temporarily removed firearms.

The RCMP is examining Van Rootselaar’s online activity, past interactions with law enforcement, and mental-health history to determine motives and planning. Archived posts show she photographed herself at a gun range, claimed to 3D-print a bullet cartridge, and engaged in discussions about firearm videos on YouTube.

The case highlights growing challenges for AI and social-media companies in balancing user privacy with public safety. OpenAI trains its models to discourage real-world harm and routes flagged content to human reviewers, who can alert authorities when there is a credible and imminent threat. The company said decisions weigh potential risks of violence against privacy concerns and the distress caused by involving law enforcement unnecessarily.

The investigation continues as authorities piece together Van Rootselaar’s digital and personal history leading up to the tragedy.