A police vehicle is parked outside a high school, the site of a deadly mass shooting in the town of Tumbler Ridge, B.C.Jennifer Gauthier/Reuters
Months before a deeply troubled teen opened fire in the Tumbler Ridge school, killing five young students and an education assistant, the shooter’s communications with an AI chatbot were flagged by OpenAI.
The shooter’s account was banned in June for violating the company’s usage policy, but the posts did not meet OpenAI’s threshold for notifying law enforcement, the company behind ChatGPT said Friday in a statement to The Globe and Mail.
The San Francisco tech company contacted the RCMP after the Feb. 10 mass shooting.
The Wall Street Journal reported Friday that employees at ChatGPT wanted law enforcement to be warned after the shooter’s posts about gun violence last June were flagged by OpenAI’s automatic review systems.
How a day of fear and grief unfolded in Tumbler Ridge
Their concerns were rebuffed, the WSJ reported, quoting unnamed sources familiar with the matter.
“Our thoughts are with everyone affected by the Tumbler Ridge tragedy,” the company said in the statement.
“We proactively reached out to the Royal Canadian Mounted Police with information on the individual and their use of ChatGPT, and we’ll continue to support their investigation.”
The RCMP continues to investigate the mass shooting, which left eight people dead, including the shooter’s mother and half-brother whose bodies were found in the family residence. The 18-year-old shooter also died of an apparent self-inflicted wound.
In its statement, OpenAI said for posts to trigger a referral to law enforcement, they must indicate “an imminent and credible risk of serious physical harm to others.”
Last June, the statement said, the company did not identify “credible or imminent planning.”
Remembering the Tumbler Ridge shooting victims: Eight lives lost
The statement went on to say that the risk of over-reporting to law enforcement can cause “distress” to a young person and family if officers show up unannounced. Such over-reporting can also “introduce unintended harm,” and prompt privacy concerns, the statement said.
It also noted that ChatGPT is trained to discourage posters from intending harm and “to avoid providing advice that could result in immediate physical harm to an individual. ”
The statement closed by saying the company regularly reviews its policies on alerting law enforcement. It did not say whether a review would be prompted by this incident.
The RCMP confirmed Friday that OpenAI contacted its investigators after the shooting.
RCMP Staff Sgt. Kris Clark said a “thorough review of the content on electronic devices, as well as social media and online activities” of the shooter is taking place.
He said “digital and physical evidence is being collected, prioritized, and methodically processed.”
Questions linger around the Tumbler Ridge shooter’s access to guns
Chris McBryan, a former RCMP member who served with the Canadian Firearms Program before retiring a decade ago, said although the shooter is dead, the RCMP will likely be working to determine whether anyone else was involved in planning.
RCMP Deputy Commissioner Dwayne McDonald said last week that of the two firearms used at the school, police are unsure of the origins of the one that caused the most damage and had never been seized by police.
“We’re trying to determine how the suspect got that firearm. The investigation is continuing.”
McBryan said figuring out where the weapons of unknown origin came from, and whether any crimes were committed by other people in relation to those weapons, would likely be a key focus of the ongoing investigation.
With a file from Joe Castaldo and The Canadian Press