The maker of ChatGPT is in hot water for not sounding the alarm on the suspected Tumbler Ridge shooter.
Driving the news: After its automated systems flagged violent conversations between ChatGPT and the alleged Tumbler Ridge, B.C., shooter, OpenAI decided not to alert Canadian law enforcement about the potential threat, despite multiple concerned employees urging senior leadership to do so as far back as June.
According to the Wall Street Journal, OpenAI did ban the account of Jesse Van Rootselaar, but said her activity didn’t meet the bar to contact law enforcement.
Van Rootselaar’s interactions with the chatbot — which described several scenarios involving gun violence — were viewed by some OpenAI staffers as showing an intent to commit real-world violence.
Why it matters: The onus has long been on tech giants like Meta and Google to report nefarious online activity to law enforcement. Now, with ~250 million people interacting with ChatGPT every day, OpenAI is finding itself in a similar position.
Zoom out: There are at least two ongoing lawsuits in the U.S. alleging that OpenAI failed to alert authorities about users who discussed acts of violence and self-harm with ChatGPT, and later acted on it.
The plaintiffs are the families of a 16-year-old who died by suicide and a man who murdered his mother after the chatbot allegedly convinced him she was conspiring to kill him.
Bottom line: Tech firms generally don’t contact law enforcement proactively with information on users, even if there is suspicious activity. Examples like this might make that seem like a bad policy, but the privacy implications of changing those rules could also be a slippery slope.—LA