British Columbia is suing OpenAI over ChatGPT warnings flagged before a mass shooting
On July 7, 2026, British Columbia's attorney general announced the province would pursue legal action against OpenAI, alleging its safety teams internally flagged the eventual Tumbler Ridge shooter's violent ChatGPT prompts months before the February 2026 attack, yet leadership did not notify police. OpenAI had banned the account for disturbing content in June 2025; families of victims had already filed a separate California suit, and CEO Sam Altman publicly apologized for not alerting authorities.
Records by entity: OpenAI
Twelve OpenAI employees reportedly urged the company to notify police about the account. No one did.
Key facts
- What
- On July 7, 2026, British Columbia's attorney general announced the province would pursue legal action against OpenAI, alleging its safety teams internally flagged the eventual Tumbler Ridge shooter's violent ChatGPT prompts months before the February 2026 attack, yet leadership did not notify police.
- Incident date
- Feb 1, 2026
- Who
- OpenAI
- Failure mode
- Brand & Safety Incident
- AI surface
- Chatbot
- Severity
- Catastrophic
What happened
British Columbia's attorney general said the province had retained legal teams in BC and California to hold OpenAI accountable for what it called a documented failure to notify law enforcement about explicit, flagged threats made on ChatGPT by the person responsible for the February 2026 Tumbler Ridge mass shooting. According to the province, OpenAI's internal reports showed safety teams flagged the shooter's violent prompts months before the attack, and the company banned the account for disturbing content in June 2025, but leadership did not alert police. Families of seven victims had already sued OpenAI and Sam Altman in California; OpenAI said in February it had considered a referral to law enforcement but decided the activity did not indicate an imminent, credible risk. Altman later published an apology.
What broke inside the model
- 01 · TriggerA user prompts the model in public view.
- 02 · Model stepThe model produces unsafe or off-brand output.
- 03 · Control gapNo filter holds the line before publish.
- 04 · FailureThe output goes public unchecked.
- 05 · ConsequenceA reputational or safety incident lands.
A contained signal crosses into output that goes public.
The model surfaced and its safety systems correctly flagged content describing and planning violent scenarios, which is the detection working. The failure was downstream: the organization's escalation policy did not translate a high-severity safety flag into a real-world notification, and an account ban closed the loop internally without triggering outside intervention. A flag that a human should have acted on sat inside the system.
What it cost
Sources
Cite this entry
https://failureindex.ai/failures/british-columbia-sues-openai-tumbler-ridge-warningsAI Failure Index. "British Columbia is suing OpenAI over ChatGPT warnings flagged before a mass shooting" (FI-0708). Realm Labs. https://failureindex.ai/failures/british-columbia-sues-openai-tumbler-ridge-warnings (indexed Jul 10, 2026).Data fields CC-BY 4.0, prose citation permitted. Incident ID FI-0708. Full dataset at /data.
Note from Realm Labs, the Index steward
How Realm would have caught this
- Prism
- OmniGuard
- AI Detection & Response (AIDR)
Realm ties a high-severity safety detection to a governed escalation policy, so a flagged trajectory routes to a defined action and owner rather than resolving silently with a ban. The detection-and-response record preserves who was notified and when, which is exactly the accountability gap this case turns on.