On Monday, September 21, Niki Sharma, Attorney General for the province of BC, announced the province had filed suit in the District of Northern California against OpenAI, related to the Tumbler Ridge tragedy.
“The people of Tumbler Ridge have endured an unimaginable loss, and they deserve answers and accountability,” says Sharma. “As part of our ongoing efforts to achieve that, British Columbia is filing a lawsuit against OpenAI in California for its failure to notify law enforcement of threats made on its platform prior to the mass shooting at Tumbler Ridge Secondary school.”
This action, says Sharma, is an important step towards justice for the families, students, educators and community members “whose lives were forever changed by the events of February 10, 2026.”
The province is working with the Board of Education for SD59 Peace River South as this matter proceeds through the courts. Indeed, the board is listed as a co-complainant.
They are bringing forward eight claims against OpenAI. These are:
Negligence (failure to warn law enforcement). “British Columbia and SD59 allege that OpenAI designed, developed, trained, tested, deployed, operated, and controlled ChatGPT, including the GPT-4o model and subsequent models, and offered ChatGPT to consumers, including the shooter,” says the filing, arguing that OpenAI owed a duty of reasonable care to avoid creating unreasonable risks of physical harm to foreseeable victims. “The duty became more acute once OpenAI had identified the shooter as a high-risk user presenting a foreseeable threat of physical harm to others.”
Members of OpenAI’s specialized safety team recognized the real-world risk of violence and recommended that OpenAI notify the RCMP. OpenAI’s leadership overruled that recommendation and warned no one, the complaint asserts.
Negligent entrustment. The province argues that OpenAI “owed a duty of reasonable care to avoid creating unreasonable risks of physical harm to foreseeable victims exposed to violence facilitated or exacerbated by ChatGPT, and a further duty to exercise reasonable care to prevent foreseeable re-access by the shooter following the deactivation of their first account.”
They also allege that OpenAI’s chats were an “interactive engagement with someone who had a propensity for violence. In June 2025, OpenAI’s system flagged their account and routed it to the specialized pipeline for users planning to harm others, and OpenAI’s trained reviewers concluded it presented a credible threat of serious physical harm to real people and recommended contacting the RCMP. OpenAI also had actual knowledge, from prior incidents, including the Las Vegas Cybertruck bombing, the Florida State University shooting, and the Finland school stabbing, that ChatGPT was being used to plan real-world attacks. Possessing that knowledge, OpenAI furnished substantial assistance to the shooter through ChatGPT’s deliberate configuration.”
Aiding and abetting a mass shooting. The shooter committed “intentional torts” (civil action causing harm). OpenAI had flagged the shooter’s account, and knew that ChatGPT had been used in previous real-world attacks, yet “made a conscious and intentional decision to continue supplying the shooter (through ChatGPT) with encouragement and information that would assist [them] in achieving their violent goals.”
Negligence (failure to warn). Different than the earlier failure to warn law enforcement, this claim argues that OpenAI’s duty to warn became more acute once OpenAI identified this shooter as a known, high-risk user. “Defendants knew these risks were not apparent to users, to the public, or to the public bodies responsible for their protection. A reasonably prudent AI company would have warned users and the public, maintained the refusal protocols that once governed violent content, and alerted authorities when its own systems identified a user engaged in planning a real-world attack.”
Negligent undertaking. Here, the province argues that OpenAI “detects users who present a risk of harm to others, routing flagged accounts to a dedicated review process for users who may be planning to harm others, and referring imminent threats of serious physical harm to law enforcement.” OpenAI should have recognized these functions as necessary to protect third persons, including the Crown and SD59 and other foreseeable victims.
Negligence (design defect). The government argues that OpenAI allegedly deliberately configured ChatGPT to maximize engagement by, among other things, weakening or removing prior requirements that the system reject dangerous premises; instructing ChatGPT to remain in conversations and be validating rather than terminate or sharply redirect conversations presenting serious risk; and prioritizing human-like mirroring of user emotions and beliefs over reliable refusal behaviors in response to violent ideation.
Strict product liability (design defect). The filing argues that a product is defectively designed “when it fails to perform as safely as an ordinary consumer would expect in reasonably foreseeable use, or when the risks of the design outweigh its benefits.” They allege that ChatGPT fails both tests.
Strict product liability (failure to warn). Here, the Crown argues that ChatGPT reached consumers without adequate warning of dangers that were neither open nor obvious; that its safety features degraded during the extended, multi-turn conversations its design encouraged; that it posed heightened, specific dangers when used by individuals experiencing violent ideation toward third parties; and that it would supply dangerous restricted content, including discussing gun violence. More damning, “The dangers were known to Defendants at the time of distribution. Well in advance of both June 2025, when OpenAI decided not to report the shooter’s account, and February 10, 2026, when the shooter attacked, others had already used ChatGPT in connection with violent public attacks.”
As a result, the province and SD59 “are seeking compensatory damages, both past and future, for the injuries the Crown and SD59 sustained as a direct and proximate result of Defendants’ conduct.
This includes seeking damages against the cost of tearing down and rebuilding the school, providing acute and ongoing health care, victim services and trauma-informed mental-health care, educational continuity and the recruitment and training of replacement staff.
They are also looking to build a community wellness center.
They are also looking for “punitive damages, restitution, and disgorgement, in an amount permitted by law, costs and expenses to the extent authorized by statute, contract, or other law and prejudgment and post-judgment interest as permitted by law.
In addition to pecuniary relief, the filing says they are looking for “injunctive relief” against the company, enforcing them to implement safeguards that reliably refuse, terminate, or de-escalate conversations in which a user expresses intent to harm others or seeks assistance with violence, weapons, or explosives directed at third parties; implement automated detection and human-review escalation of conversations indicating that a user may be planning to harm others, and prohibit the product from proposing or accepting fictional, roleplay, “educational,” or “research” framings as a means of producing otherwise-restricted violence, weapons, or explosives content; establish and adhere to a policy of referring to law enforcement any user whose conversations present an imminent or credible threat of serious physical harm to others, consistent with OpenAI’s own stated escalation criteria; display clear, prominent warnings disclosing that the product’s safety features degrade during extended, multi-turn conversations, and that the product may generate detailed violent, weapons, or explosives content; implement effective, durable account-level enforcement so that a user banned for violent misuse cannot promptly re-register and regain access to the product on the same identity, devices, or payment method; adopt and enforce an effective red-line principle that the product will not facilitate acts of violence, and cease marketing the product without appropriate safety disclosures regarding its capacity to generate content facilitating harm to others; implement auditable controls regarding safety training and safeguards; and submit to quarterly compliance audits by an independent monitor.”
Finally, they are looking for OpenAI to cover the Crown and SD59 reasonable attorneys’ fees, experts’ fees, and costs of litigation.
In addition to going after OpenAI, the province is also going after the federal government, asking them to improve policy around amendments to the Canadian Criminal Code. “This case highlights the urgent need for strong national safeguards for artificial intelligence technologies and online platforms,” says Sharma. “I have written to federal ministers Fraser, Solomon and Miller with clear policy ideas to amend the Criminal Code of Canada to ensure a pathway to human accountability for AI’s actions. As technology evolves, our laws and safety frameworks must keep pace. British Columbians deserve confidence that credible threats of violence will be treated seriously, and that those who profit from offering AI services to the public are held accountable for any criminal actions that can appropriately be traced back to them.
“To date, OpenAI has not taken meaningful steps to address the concerns raised by survivors, families and the Tumbler Ridge community. The company still has an opportunity to demonstrate accountability by strengthening safeguards, improving transparency and taking concrete action to help prevent similar tragedies in the future.
“Our government remains committed to supporting the Tumbler Ridge community and to working with partners at all levels to strengthen safeguards that help keep people safe online and in their communities.
“This action today is consistent with the province’s long-standing commitment to hold powerful corporations accountable when their products or practices cause harm to people and communities. From tobacco and opioid manufacturers to vaping companies, the Government of British Columbia has used the courts to pursue accountability and ensure that people in BC are not left bearing the costs of corporate wrongdoing
“Legal proceedings cannot undo the immense loss experienced by the people of Tumbler Ridge. However, holding OpenAI accountable, seeking much needed answers and demanding stronger protections are important steps for preventing future tragedies.”
Trent is the publisher of Tumbler RidgeLines.
