In the months leading up to the tragic mass shooting in Tumbler Ridge, British Columbia, Jesse Van Rootselaar, the identified suspect, had been causing concern among employees at OpenAI. Back in June, Rootselaar engaged in conversations with ChatGPT that included descriptions of gun violence, prompting the chatbot’s automated review system to take notice. Several staff members expressed worries that these interactions might foreshadow actual violent actions and advised company leaders to notify authorities. However, the company ultimately chose not to proceed with such a referral.

Kayla Wood, a spokesperson for OpenAI, explained to The Verge that while the option to report the account to law enforcement was considered, it was concluded that the situation did not represent an “imminent and credible risk” to others. Wood noted that an examination of the logs did not reveal any active or immediate plans for violence. Although OpenAI decided to ban Rootselaar’s account, no additional preventive measures were implemented.

Wood stated, “Our thoughts are with everyone impacted by the Tumbler Ridge tragedy. We proactively contacted the Royal Canadian Mounted Police to share information about the individual’s interactions with ChatGPT, and we remain committed to assisting their investigation.”

On February 10th, a devastating mass shooting occurred, resulting in the deaths of nine people and injuries to 27 others, including Rootselaar. This incident, the deadliest in Canada since 2020, took place primarily at Tumbler Ridge Secondary School, where Rootselaar was discovered deceased from an apparent self-inflicted gunshot wound.

In hindsight, the decision not to alert law enforcement may seem questionable, but Wood emphasized that OpenAI strives to balance privacy and safety, aiming to avoid the potential negative consequences of indiscriminate law enforcement referrals.

Updated February 21st: A statement from OpenAI has been added to this article.