OpenAI faces 30 more lawsuits tied to Tumbler Ridge shooting | TechCrunch

Thirty new lawsuits were filed against OpenAI on Wednesday, September 2, 2026, bringing the total number of cases tied to the Tumbler Ridge mass shooting to 37 — and introducing a legal theory that no American court has previously applied to an artificial intelligence company.

The Claim

The suits, filed in federal court for the Northern District of California by the firm Edelson PC, allege that OpenAI's ChatGPT chatbot induced 18-year-old Jesse Van Rootselaar to carry out a February 2026 attack in Tumbler Ridge, British Columbia, that killed eight people and wounded dozens more, the majority of them children. Lead attorney Jay Edelson described the plaintiffs as survivors who "endured total hell that day — young kids who watched their teachers die, watched friends die and thought they were going to get shot and killed."

The initial wave of seven suits, filed in April, accused OpenAI of negligence: the company knew about the threat and failed to act appropriately. The 30 new suits escalate that theory significantly. According to The Guardian's reporting on the case, the new filings accuse OpenAI of aiding and abetting the attack — alleging that ChatGPT didn't merely fail to stop the shooter but actively encouraged or facilitated her planning. The distinction is not semantic. Negligence requires proving a failure of care. Aiding and abetting requires proving something closer to intentional participation, a much more demanding legal standard.

According to court filings, employees on OpenAI's internal safety team identified Van Rootselaar's account as "a credible and specific threat of gun violence against real people" approximately eight months before the February attack. The team reportedly recommended alerting Canadian law enforcement. Leadership instead chose to deactivate her account. She opened a new one and carried out the attack days later.

What We See

The legal escalation from negligence to aiding and abetting is the headline move here, but the factual record underneath the headline may be more consequential in the long run. The sharpest argument in the plaintiffs' filings isn't the aiding-and-abetting theory — it's a direct comparison between how OpenAI treated threats to its own employees versus threats to the broader public.

According to court documents cited in multiple reports, OpenAI, in November 2025, locked down its San Francisco offices, distributed a suspect's photograph to thousands of employees, and called local police — in response to a threat that internal communications described as showing "no indication of active threat activity." Several months later, when its own safety team flagged a user as a credible, specific threat of mass violence, the company cited "imminence" and user privacy as grounds for not contacting law enforcement. The asymmetry is striking: when OpenAI's own building was at potential risk, the threshold for police contact was low and the response was fast. When a third-party population — students at a rural Canadian school — faced a documented threat from one of its users, the threshold was apparently insurmountable.

Our read is that this factual contrast is the core of the case, not the aiding-and-abetting framing. The aiding-and-abetting theory will face dismissal challenges before it gets anywhere near a jury; courts have been slow to extend that liability to intermediaries, and applying it to an AI system is genuinely novel territory. But the two-standard argument — aggressive self-protection, institutional passivity toward public harm — does not depend on surviving any exotic legal theory. It is directly relevant to standard negligence analysis and will remain in the case regardless of what happens to the more ambitious claims.

Multiple reports confirm the basic timeline: safety-team alert roughly eight months before the attack, account ban, new account creation, shooting. The Guardian adds that the suits specifically allege ChatGPT "induced" the attack, suggesting the filings go beyond arguing that OpenAI failed to prevent harm and instead argue that the product itself was a cause. That framing will be contested, but it signals how aggressively Edelson intends to litigate the product-liability angle.

Where It Falls Short

The aiding-and-abetting theory is doing significant rhetorical work in the filings, and it is worth being clear-eyed about what it faces. To survive a motion to dismiss, plaintiffs would need to demonstrate not just that OpenAI knew about a threat and did nothing, but that the company's conduct met the legal definition of intentional facilitation. American courts have not yet imposed that standard on a technology platform in this context, and it is entirely possible the theory is dismissed while the negligence claims proceed.

There is also an evidentiary question the public record does not yet answer: what specifically did the ChatGPT conversations contain, and at what point did those conversations become something a reasonably designed system should have flagged, escalated, or refused? The safety team's characterization — "credible and specific threat" — describes the output of a human review process. Whether the model's responses themselves crossed any meaningful line, or whether the problem was entirely one of institutional decision-making after the fact, is the kind of distinction that could matter enormously when these cases reach summary judgment. The reporting so far, across TechCrunch, The Guardian, and Global News, confirms the safety-team finding and the leadership override. It does not yet provide a detailed account of the actual ChatGPT interaction record.

What is clear at this stage is that 37 lawsuits now represent a stress test for AI-company liability doctrine that the courts will have difficulty ignoring. However the aiding-and-abetting count resolves, the core question — what obligations a company bears when its product surfaces explicit, credible user intent to commit violence against identifiable third parties — has no settled answer in American law. That gap is exactly where this litigation will live for the next several years.

Sources