OpenAI faces 30 new Tumbler Ridge suits, including aiding and abetting
Two stories today pull in opposite directions on the same question: who gets to decide how much risk an AI company has to absorb. In California, families from a Canadian school shooting escalated from negligence to something far harder to prove. In New York, a mayor decided the answer for 600,000 children is simply: not yet.
Thirty new lawsuits filed Tuesday in California federal court accuse OpenAI of "aiding and abetting" the Tumbler Ridge shooting — the first time the plaintiffs have gone beyond claims of negligence. The filings, led by attorney Jay Edelson, target the February 10 attack in British Columbia in which teenager Jesse Van Rootselaar killed her mother and half-brother at home, then six more people at Tumbler Ridge Secondary School before dying by suicide. The complaints allege OpenAI gave the shooter "substantial assistance and encouragement," and that the company kept access open by deactivating the account rather than issuing a system-wide ban — which they say let her simply sign back up with another email. OpenAI has said the activity didn't meet its internal bar for an "imminent and credible risk" serious enough to involve law enforcement.
The sharper allegation is structural: the complaints claim OpenAI's intelligence and investigations team — the group that flags violent users — was placed under Chris Lehane, the company's political operator, so the call on whether to alert Canadian police was made on PR grounds rather than threat-assessment grounds. One filing notes that when OpenAI itself felt threatened, it locked down its offices and called San Francisco police without waiting for anything to become imminent. OpenAI chief strategy officer Jason Kwon called the Lehane claim "absolutely false." The aiding-and-abetting theory needs proof of intent, which is a steep hill, and these claims look built for discovery rather than for a clean win — the real prize is internal documents about how referral decisions actually get made.
New York City will bar students in 2-K through eighth grade from using AI in classrooms for the 2026-2027 school year, a one-year moratorium covering roughly 600,000 public school students. Mayor Zohran Mamdani's Department of Education policy also bans companion chatbots in every grade, stops teachers from using AI to grade work, limits individual screens until third grade, and caps middle schoolers at 45 minutes of screen time a day. High schoolers get AI only in limited cases, plus AI literacy classes twice a year. Exceptions exist for students with disabilities, multilingual learners, and computer science coursework.
What's notable is how deliberate the carve-outs are: teachers can still use AI for lesson planning and translation, and a small pilot in a handful of high school classes will test vetted tools with hard time limits. That's not a ban on the technology — it's a ban on the technology touching children, which is a distinction most jurisdictions have so far refused to make. Districts across the country will be watching whether 600,000 students offline for a year produces measurably better outcomes, or just a very expensive experiment in holding a line.
Amazon says its shopping assistant can now "definitively" tell you whether a message claiming to be from Amazon is real. Alexa for Shopping checks sender information, content, timing, and metadata against the billions of messages Amazon has actually sent, and the company says it improves as customers report more suspicious messages. It follows earlier attempts at the same problem, including an email address that anyone can forward suspicious messages to for a verdict.
It's a small feature with an unusually clean theory behind it: Amazon holds the ground truth for its own outbound mail, so it's one of the few companies that can answer the question with something better than a guess. Scam messages impersonating retailers are a volume game, and shifting the verification into the assistant removes the step where people get fooled — deciding whether to trust the email at all.
What to watch: whether the California court lets the aiding-and-abetting claims survive a motion to dismiss. That ruling will tell the industry more about AI liability than any of the negligence suits have.
Should a lab that flags a violent user be able to weigh PR risk in the decision to call police? Tell us in the comments.
