British Columbia is seeking compensation from OpenAI after a school shooting in Tumbler Ridge, arguing that the company’s handling of violent ChatGPT conversations may have contributed to a tragedy whose costs will extend far beyond the day of the attack.
The province filed suit over the February 10, 2026 shooting at Tumbler Ridge Secondary School, where eight people were killed. The victims included five children and an education assistant at the school. The shooter’s mother and half-brother were also killed before the attack.
The case is not simply about what happened inside a chatbot. It is about who should pay when warnings appear in a digital conversation, safety systems identify danger, and intervention does not follow.
A community facing years of recovery
According to the complaint, the school was demolished after the attack. British Columbia says its response has involved far more than replacing a building. The province is seeking compensation for temporary schooling, replacement educators, policing, health services, counseling, victim support and long-term recovery programs.
Ars Technica reported that the lawsuit also asks a court to order ChatGPT to terminate violent conversations automatically. That request would move beyond voluntary safety policies and impose a legal requirement on how the service responds to users discussing potential violence.
The province’s central demand is access to the shooter’s ChatGPT logs. OpenAI has provided the records to the Royal Canadian Mounted Police, but has not released them publicly. The company has cited privacy concerns, creating a conflict between protecting personal information and allowing a community to understand whether the system played a role in the attack.
The logs could help answer several unresolved questions. Did ChatGPT reinforce violent ideas? Did the conversations contain a credible and specific threat? Were OpenAI’s safety systems able to recognize the danger? And if they did, why did the company allegedly stop short of contacting law enforcement?
A warning that did not reach police
The complaint, as described in the report, says OpenAI detected the conversations in June 2025. Human reviewers reportedly judged the user to be a credible and specific risk. The account was deactivated, but the matter was allegedly not referred to police. The user then reportedly created another account.
That sequence is likely to become one of the most important issues in the case. Account removal can prevent a person from using a service, but it does not necessarily alert people who may be able to respond to a threat in the physical world. The lawsuit argues that OpenAI’s public statements about routing serious threats to trained reviewers and law enforcement may have created responsibilities beyond the company’s contractual terms.
OpenAI has not been found liable, and the allegations remain unproven. The company’s decisions, the meaning of the conversations and the extent of any connection to the shooting will be contested in court.
A test for the chatbot industry
The lawsuit could become a landmark dispute over whether conversational AI companies have a duty to warn when violent planning develops over time. It may also test how courts weigh privacy against public safety when a platform holds potentially crucial evidence.
Chatbots are designed to keep conversations going, often across many exchanges. That design creates a difficult problem when a user moves gradually from disturbing ideas to concrete planning. A single message may seem ambiguous, while a longer conversation can reveal a pattern.
British Columbia’s claim asks whether companies must act on that pattern, and whether failing to do so can make them financially responsible for the consequences. The answer could shape product design, emergency reporting policies and the limits of corporate responsibility across the generative AI industry.
This article was written with the assistance of an AI system and published automatically.