British Columbia Says It Will Sue OpenAI Over Tumbler Ridge Warning Claims

British Columbia’s OpenAI lawsuit draws scrutiny of when chatbots must warn police

British Columbia’s September lawsuit against OpenAI and CEO Sam Altman alleges they knew about conversations preceding the Tumbler Ridge shooting and failed to prevent it. An October 17 Al Jazeera examination puts that accusation alongside a difficult question: how should chatbot companies distinguish private expressions of violence from threats requiring a police warning?

The provincial government’s civil case concerns the February 10, 2026, attack in British Columbia. Jesse Van Rootselaar killed his mother and half-brother at home, then killed six people at a school, including five students, before dying by suicide. Victims’ families have also brought separate lawsuits against OpenAI.

Al Jazeera, citing earlier reports, says Van Rootselaar consulted ChatGPT about carrying out mass murder. OpenAI’s leadership publicly apologized for not notifying authorities. The company did not respond to Al Jazeera’s request for comment on the lawsuits. The province’s accusations remain allegations, not a court finding that OpenAI caused the attack.

The technical problem starts before any police referral: deciding what a conversation means. Fazl Barez, an Oxford researcher specializing in AI safety and governance, told Al Jazeera that current models cannot reliably distinguish hypothetical exchanges from those likely to lead to harmful behavior. A disturbing subject alone does not establish intent.

Barez gave two examples: a fiction writer exploring a murder plot and a graduate student studying dangerous pathogens. Both could have legitimate reasons for discussing material that looks alarming in isolation. Setting the reporting threshold too low could flag those users despite the educational or creative context of their questions.

His proposed distinction depends on context, intent and patterns across conversations. Someone seeking support for intrusive violent thoughts may sound different over time from someone gradually revealing an attack plan. Assessing a credible threat may therefore require examining multiple exchanges rather than treating one statement as decisive, Barez said.

Forcing companies to implement a surveillance system implicates very serious civil liberties concerns, and that’s probably a bridge too far for me

A familiar comparison is the duty some mental-health professionals have to warn identifiable people about credible threats. Fordham law professor Chinmayi Sharma cautioned that therapists bring training, patient knowledge and enforceable professional standards to those judgments. AI companies do not necessarily have equivalent expertise or accountability, she told Al Jazeera.

Discussing US law, University of Colorado Boulder law professor Bryan Choi said existing case law appears unlikely to impose the same warning duty on OpenAI. A chatbot is not a licensed therapist, he said, and the company lacks a special relationship with victims. That assessment concerns US legal principles, not a ruling on British Columbia’s allegations.

Choi said negligence and product-liability laws could potentially apply, although establishing responsibility would be difficult. Those routes are distinct from automatically giving a chatbot company the duties of a clinician.

The costs of broader monitoring would reach beyond prospective attackers. Younes warned that people might avoid chatbots if disturbing thoughts could trigger disclosure to police. Sharma raised a separate concern: systems trained to detect dangerous speech could reproduce biases, disproportionately flagging some minorities while overlooking comparable speech from others.

Sharma proposed independent institutions to develop threat-assessment procedures and explore sharing only relevant information while protecting personal details. Younes called for clear notice about monitoring and police disclosures. Neither proposal settles the lawsuit, but each addresses who should make these decisions—and what users should know before confiding in a chatbot.