Product1 distinct publisher3 min readPublished
Adding an intent claim on top of the seven suits filed in April pushes discovery past the model and into the reporting line that decided not to call Canadian police, which is a record most teams shipping chat products do not keep.
The Product Desk · Product desk
Compiled by The Product DeskSomething wrong?How this is made
Van Rootselaar's account was deactivated, and she opened another one shortly afterward [9]. A ban is supposed to remove a user, but here the record closed and the person came back.
Negligence and intent are different discovery problems. The seven April complaints plus the 30 being filed this week make 37 suits from one firm over one incident [18], and the new ones allege OpenAI aided and abetted the attack rather than negligently failing to stop it [4]. OpenAI was already defending suits that put product design itself in issue, alleging ChatGPT's design contributed to users' suicides, violent acts and severe mental health crises [17].
Intent claims run through people, so the new filings read like an org chart. They name Chief Global Affairs Officer Chris Lehane as the person who told staff to stand down, offer no direct evidence he was personally involved, and OpenAI has denied his involvement [12]. One complaint asserts that the Intelligence and Investigations Team, described as the only team inside OpenAI responsible for identifying users who pose a threat of real-world violence, was placed under Lehane's control, and that the decision was made by him or someone in his chain of command and ratified by Sam Altman, on information and belief [13]. Lead lawyer Jay Edelson told TechCrunch the firm is basing this in part on the org chart and its broader investigation, and that Lehane and Altman will be critical witnesses [15]. TechCrunch could not confirm that Lehane had authority to override the team, or that he was involved in this case at all [16]. He is not a defendant. Altman is [14].
OpenAI's answer is a threshold: the activity did not meet its internal bar for an imminent and credible risk of serious physical harm, the trigger required to engage law enforcement [10]. Chief strategy officer Jason Kwon, who oversees the human reviewers and the legal department, said the judgment "is not infallible, but it is always rooted in looking out for this balance for people" [11]. That threshold will be read aloud beside the reporting, from the Wall Street Journal, that staff were alarmed by conversations about gun violence and advice on planning an attack [7] and urged leaders to alert Canadian police before leaders decided against it [8].
The grid worth drawing on your own product has two axes: whether a detection creates a durable record outside the product team, and whether the response binds to a person or only to a session. Only the record-plus-person-binding quadrant lets you show both what you decided and that it held. Record plus account-only means you've at least documented the gap you chose to leave, uncomfortable but survivable. With person-binding and no record, you acted, but you can't prove it. Neither means your reporting line answers the question for you, which is the posture these complaints are built to exploit.
The bar, who may override it, and what happens when a deactivated account reappears within a week are the things worth having in writing. A written bar can be read back to you every time it failed, but an unwritten one is discoverable too, in the shape of whichever executive's chain of command happened to contain the reviewers.
Ranked by verification strength, evidence, and original report placement.
Edelson PC filed seven lawsuits against OpenAI in April on behalf of victims and families connected to the Tumbler Ridge mass shooting.
Edelson PC is filing 30 additional complaints against OpenAI this week over the Tumbler Ridge shooting.
The new plaintiffs include teachers, a principal, and students who were in the building during the attack but were not physically shot.
The new filings for the first time accuse OpenAI of aiding and abetting the mass shooting rather than just negligently failing to prevent it; that claim requires proving intent from OpenAI and is likely to face early dismissal challenges.
The fresh wave of lawsuits was filed in a California court on Tuesday.
The shooting occurred in Tumbler Ridge, British Columbia on February 10, when teenager Jesse Van Rootselaar killed her mother and half-brother at home, then killed six more people at Tumbler Ridge Secondary School and injured dozens before dying by suicide.
Distinct publishers with included, body-backed reporting in this cluster.
1 article · September 2, 2026
Follow any of these and your For You feed starts watching them — no settings page required.
product
Siting Is Now a Schedule Risk: 75 Data Centre Projects, $130bn, One Quarter1 distinct publisher
product
OpenAI's CFO calls an IPO "another fundraise" while the model calendar slips2 distinct publishers
build
Hugging Face's $13B process puts most teams' model pipeline under a single owner2 distinct publishers
security
OpenAI's 13-17 tier turns teen AI safety into an age-assurance problem1 distinct publisher
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
One newsroom holding the complaints, and a candid gap at the centre
Everything verifiable here is documentary: TechCrunch has the complaints, an on-record statement from Jason Kwon, and the lead plaintiffs' lawyer speaking directly. What the reporting cannot stand behind is the claim that carries the story — that Chris Lehane ordered the stand-down — which the filings themselves plead on information and belief, and which TechCrunch says outright it could not confirm. The chat-log facts are relayed from the Wall Street Journal rather than independently obtained, so the factual spine and the accusation sit at very different strengths.
Not an uptake story
Nothing in this reporting measures usage, deployment, or third-party behaviour. The only quantities are plaintiff counts, and a growing docket is not adoption of anything. No policy change at OpenAI, no regulatory action, and no move by another provider is documented, so we leave this unscored rather than dress up the filing tally as traction.
The loudest count is the thinnest one
The gap lives in the filings, not the coverage. Aiding and abetting requires intent, and the support offered for it is an org chart plus an unspecified "broader investigation" that Edelson is explicitly holding back. Naming an executive who is not a defendant sharpens the narrative at no procedural cost. TechCrunch narrows the gap by printing the pleading standard, the denial, and its own failure to verify — which is why this reads as overstated advocacy rather than overstated journalism.
Two adversaries, both talking for the record
Every substantive assertion here comes from a party with something at stake in it. A plaintiffs' firm expanding from seven complaints to 37, and telling a reporter it is withholding evidence, benefits from a named villain with a spin-doctor résumé. OpenAI, already defending suits over suicides and violent acts and mid-fallout from a sandbox escape, benefits from a denial that isolates the referral decision inside its trained reviewers. Neither incentive makes either side wrong; it does mean the org chart question needs a source who is not a litigant.
Solid on the paperwork, open on the reporting line
We are confident about what was filed, by whom, when, and what OpenAI said in response — that is all documented and quoted. We are not confident about the fact the story turns on. One outlet, one plaintiff-side document set, an unverified chain of command, and no second newsroom testing it yet: enough to report, not enough to conclude.