Associated Incidents
Months before Jesse Van Rootselaar became the suspect in the mass shooting that devastated a rural town in British Columbia, Canada, OpenAI considered alerting law enforcement about her interactions with its ChatGPT chatbot, the company said.
While using ChatGPT last June, Van Rootselaar described scenarios involving gun violence over the course of several days, according to people familiar with the matter.
Her posts, flagged by an automated review system, alarmed employees at OpenAI. Internally, about a dozen staffers debated whether to take action on Van Rootselaar's posts. Some employees interpreted Van Rootselaar's writings as an indication of potential real-world violence, and urged leaders to alert Canadian law enforcement about her behavior, the people familiar with the matter said.
OpenAI leaders ultimately decided not to contact authorities.
A spokeswoman for OpenAI said the company banned Van Rootselaar's account but determined that her activity didn't meet the criteria for reporting to law enforcement, which would have required that it constituted a credible and imminent risk of serious physical harm to others.
On Feb. 10, Van Rootselaar was found dead from what appeared to be a self-inflicted injury at the school scene of a mass shooting that killed eight people and left at least 25 injured. The Royal Canadian Mounted Police identified Van Rootselaar, an 18-year-old trans woman, as the suspect.
The company reached out to the RCMP after it learned of the shooting and is supporting its investigation, the spokeswoman said.
"Our thoughts are with everyone affected by the Tumbler Ridge tragedy," the company said in a statement.
Other aspects of Van Rootselaar's digital footprint emerged in the days after the attack, including a videogame she created on the Roblox platform that simulated a mass shooting. On social media, the suspect shared her concerns about the process of transitioning and her interests in anime cartoons and illicit drugs.
Online platforms have long debated how to balance questions of privacy for their users with public safety in their decisions to alert certain users to law enforcement. That debate is now coming for the AI companies that power the chatbots to which people are confiding the most intimate details of their private thoughts and lives.
OpenAI said it trains its models to discourage users from committing real-world harm, and routes conversations in which users express intent of harm to human reviewers, who are able to refer them to law enforcement in cases where they are found to pose an imminent risk of serious physical harm.
The company said it weighs the risk of violence against privacy considerations and the potential distress caused to individuals and families by getting police involved unnecessarily.
Van Rootselaar was already known to local police before the shooting. They visited where she lived multiple times to handle mental-health concerns, and temporarily removed guns from the residence.
A specialized team of investigators has also been combing through her online activity and digital footprint for clues about the mass shooting, as well as reviewing her past interactions with police and mental-health professionals, according to RCMP Commissioner Dwayne McDonald.
Archived social-media posts show Van Rootselaar posted pictures of herself shooting at a gun range, claimed to have created a bullet cartridge using a 3-D printer and engaged in online discussion about YouTube videos made by gun enthusiasts.