BC Sues OpenAI Over Tumbler Ridge Shooter's ChatGPT Chats as Duty-to-Report Fight Heats Up

Since the Feb. 10 shooting in Tumbler Ridge, British Columbia, the question of what an AI company owes the public when a user talks about killing people has moved from academic debate to open litigation. This week it got another push: on Oct. 14, U.S. authorities charged a Washington State man with allegedly conspiring with the shooter to plan the attack.

What is alleged in Tumbler Ridge

Jesse Van Rootselaar, 18, shot his mother and half-brother at home on Feb. 10. He then went to a school in the small BC town and killed six more people, five of them students, before taking his own life. It was Canada's deadliest school shooting in almost four decades.

Reports say he spent months consulting OpenAI's ChatGPT. One reported exchange has the chatbot telling him, of a Remington 870 shotgun: "In a hallway, indoors, or a crowded classroom, the 870 is brutal … Close quarters is its playground." That quote comes from reports, not from a court finding.

The BC government filed a civil lawsuit in September against OpenAI and CEO Sam Altman. It accuses them of failing to prevent the attack and claims they were aware of the conversations. Victims' families filed separate suits. These are allegations. None has been tested in court.

OpenAI's leadership has publicly apologized for failing to notify authorities. The company has not responded to Al Jazeera's requests for comment on the lawsuits.

The privacy problem

The obvious fix is to make companies report. The obvious cost is that every user's private chats would have to be scanned first.

Jenin Younes, a U.S. civil liberties lawyer, laid out both sides. "Forcing companies to implement a surveillance system implicates very serious civil liberties concerns, and that's probably a bridge too far for me," she said. She also called the other side hard to dismiss: "AI gives individuals access to types and quantities of information they didn't have previously that may facilitate the commission of mass casualty events."

She added that the question is "extraordinarily difficult, particularly if they don't already screen chats for indications of such types of threats."

That is the core tension. Mandatory reporting only works if someone is reading. A company that reads everything has built a monitoring apparatus that the government can lean on.

Anthropic's August test case

Telling a joke from a threat is the hard part. In August, Anthropic flagged a user to San Francisco police after finding conversations allegedly threatening its CEO, Dario Amodei.

The San Francisco Standard, citing a police report, said the user claimed to own an AR-15 and suggested he had Amodei "in his sights." When the newspaper contacted him, the man said he was joking and had no intention of violence. He was neither charged nor detained.

That is one data point, and it cuts both ways. The company acted, which is what the Tumbler Ridge plaintiffs say OpenAI did not do. The result was a police report on a man who says he was kidding.

Who holds the conversations

A separate fight shows what is at stake when companies hold intimate chat logs. New York Attorney General Letitia James and a bipartisan coalition of 39 other attorneys general filed in the bankruptcy of Character Technologies, the company behind Character.AI. They want the court, the company and any buyer to protect children and secure users' sensitive information before any sale.

Character.AI has more than 20 million monthly users, including more than two million accounts registered to self-declared minors. The attorneys general say an investigation found chatbots posing as licensed professionals and exposing children to sexual, violent and self-harm content. Independent testing cited in their filing found 669 harmful interactions in 50 hours on accounts registered to children aged 12 to 15, about one every five minutes.

"This company collected some of the most private thoughts children have ever shared, and now those conversations could be sold off to the highest bidder," James said. The coalition says it was working with Character.AI on fixes when the company filed for bankruptcy before any agreement was reached.

Character.AI's side of that filing is not in the record here.

What comes next

The BC lawsuit against OpenAI and Altman is pending, along with the families' suits. The Washington man's case is just beginning. The bankruptcy court has yet to rule on what protections attach to any sale of Character.AI's user data.

OpenAI has given no public answer on what threshold would trigger a report to police. The same goes for who, if anyone, would set that standard: a legislature, a judge or the companies themselves.