OpenAI staff debated alerting Canadian police about violent ChatGPT logs months before a deadly school shooting

Over several days in June 2025, Van Rootselaar described gun violence scenarios in ChatGPT. OpenAI says it trains its models to steer users away from real-world violence. When users express intent to harm, those conversations get flagged for human reviewers, who can bring in law enforcement if there's an immediate risk of serious physical harm.

AI companies face a growing tension between privacy and public safety

According to the report, about a dozen OpenAI employees discussed the posts internally. Some saw the messages as red flags for possible real-world violence and urged senior management to contact Canadian police.

OpenAI decided not to act. A company spokeswoman told the Wall Street Journal that while Van Rootselaar's account had been suspended, her activity didn't meet the bar for reporting to law enforcement. That would have required a "credible and imminent risk of serious physical harm to others." The company contacted the RCMP after learning about the attack and is now cooperating with their investigation.

ChatGPT wasn't the only place Van Rootselaar left digital warning signs. On the gaming platform Roblox, she allegedly simulated a mass shooting in a shopping mall and joined online discussions about YouTube videos from gun enthusiasts.

On February 10, Van Rootselaar was found dead at the scene of a shooting rampage, apparently from a self-inflicted injury. Eight people were killed and at least 25 wounded. The Royal Canadian Mounted Police identified the 18-year-old as the suspect.