Lawsuits say OpenAI failed to report violent ChatGPT user before shooting

In short: Seven lawsuits filed in California claim OpenAI did not alert police about a ChatGPT user flagged internally as a credible threat before a Canadian school shooting.

What happened

Seven lawsuits were filed Wednesday in a California court over a school shooting in Tumbler Ridge, Canada. The suits were brought by families of people killed and by one mother whose child was badly injured.

The lawsuits allege OpenAI’s safety staff flagged a ChatGPT account linked to the shooter more than eight months before the attack. The suits say the safety team recommended contacting law enforcement, but company leaders chose not to. Instead, OpenAI allegedly deactivated the account and then provided instructions on how to sign up again with a different email address.

OpenAI CEO Sam Altman has apologized publicly to the community, saying the company should have alerted police about the account that was banned. OpenAI said it has a “zero-tolerance policy” for using its tools to assist violence and that it has strengthened safeguards. These include better handling of signs of distress, clearer escalation of potential threats, and improved detection of repeat rule breakers.

Why it matters

This case raises a basic question about responsibility. If a company’s system spots what looks like a real and specific threat, should it warn police, even if that means sharing user information. It also highlights a practical issue, which is that banning an account only works if the person cannot easily return, like changing the locks but then leaving a spare key under the mat.

Source: Arstechnica