Sam Altman apologizes after OpenAI failed to flag user who carried out Canada family shooting

Sam Altman, the head of OpenAI, has publicly apologized after the company acknowledged it failed to alert Canadian authorities about a user who later carried out a deadly shooting targeting her own family. The incident, which shocked the small community of Langley, British Columbia, has drawn global attention to the growing dilemma tech companies face when their systems detect signs of potential violence.

According to OpenAI, the user-identified in reports as 28-year-old Naomi Van Rootselaar-had her ChatGPT account suspended in June 2025 after automated systems and human reviewers flagged her for engaging in content related to the 'furtherance of violent activities.' While the company took internal action by cutting off access, it did not report her to law enforcement, citing current policies that prioritize user privacy unless there is a clear, imminent threat.

Van Rootselaar is accused of fatally shooting four family members during a gathering on April 20, 2026. Investigators say evidence recovered from her devices shows extensive interactions with AI tools in the months leading up to the attack, including detailed planning and ideation around violence. Some exchanges with ChatGPT reportedly included questions about weapon access, psychological manipulation, and methods to avoid detection.

In a statement released Friday, Altman said, 'We made a judgment call that did not go far enough. While we acted to disable the account, we should have done more. We are deeply sorry for the harm that followed and are committed to learning from this tragedy.' He added that OpenAI is now working with legal and ethics experts to reassess when and how it should escalate concerns to authorities.

The case has sparked a broader conversation about the role of AI platforms in public safety. Unlike social media companies, which have developed threat-reporting pipelines in coordination with law enforcement, AI firms like OpenAI operate in a less defined space. There are no standardized rules for when a pattern of concerning queries crosses the line into reportable behavior.

Privacy advocates warn against turning AI companies into surveillance arms of the state, cautioning that overreporting could chill free expression and disproportionately impact vulnerable users. But others, including some lawmakers and victim advocates, argue that when clear warning signs emerge, companies have a moral if not legal obligation to act.

Canadian officials have not yet commented on whether a report from OpenAI could have changed the outcome. Meanwhile, the company says it's launching an independent review of its safety protocols and will publish its findings later this year. For now, the tragedy underscores a growing tension: as AI becomes more deeply woven into daily life, the line between private conversation and public risk is getting harder to define.