1 hour ago
5
OpenAI CEO Sam Altman has apologized to members of a Canadian assemblage wherever a wide shooting took spot earlier this twelvemonth for not flagging the ChatGPT relationship of the shooter to instrumentality enforcement.
"The symptom your assemblage has endured is unimaginable," Altman wrote successful a missive shared Friday connected societal media by the British Columbia Premier David Eby. "I person been reasoning of you often implicit the past fewer months."
Eight radical were killed successful the Feb. 10 massacre successful the tiny assemblage of Tumbler Ridge successful northeast British Columbia. Six radical were fatally changeable erstwhile 18-year-old Jesse Van Rootselaar opened occurrence astatine Tumbler Ridge Secondary School, authorities said, and the shooter's parent and 11-year-old member were killed astatine a adjacent residence. Van Rootselaar died of a self-inflicted gunshot wound, officials said.
Altman wrote successful the letter, dated Thursday, that Van Rootselaar's ChatGPT relationship had been banned successful June 2025 — astir 8 months anterior to the shooting.
"I americium profoundly atrocious that we did not alert instrumentality enforcement to the relationship that was banned successful June," Altman said.
In February, OpenAI told CBS News that Van Rootselaar's relationship had been flagged past twelvemonth by automated maltreatment detection tools and quality investigators that place imaginable misuses of ChatGPT for convulsive activities. OpenAI said the relationship was past banned for violating its usage policies.
OpenAI said that the institution had weighed whether to emblem the relationship to instrumentality enforcement, but had determined astatine the clip that it did not airs an imminent and credible hazard of superior carnal harm to others, failing to conscionable the threshold for referral.
"Our thoughts are with everyone affected by the Tumbler Ridge tragedy," OpenAI said successful a connection to CBS News successful February pursuing the shooting. "We proactively reached retired to the Royal Canadian Mounted Police with accusation connected the idiosyncratic and their usage of ChatGPT, and we'll proceed to enactment their investigation."
OpenAI says that ChatGPT is trained to discourage real-world harm, and is instructed to garbage to assistance erstwhile it detects an illicit intent. Users that bespeak plans to harm others are flagged to quality reviewers who find whether a lawsuit poses an imminent menace of carnal harm and should beryllium referred to instrumentality enforcement, according to the company.
Altman wrote successful his missive that OpenAI volition stay focused connected preventative efforts "to assistance guarantee thing similar this ne'er happens again."
"I privation to explicit my deepest condolences to the full community," Altman said. "No 1 should ever person to endure a calamity similar this."
Earlier this week, Florida Attorney General James Uthmeier announced a transgression probe into OpenAI aft reviewing messages betwixt ChatGPT and a Florida State University pupil accused successful an April 2025 field shooting that killed 2 radical and wounded respective others.
Uthmeier said his squad determined that ChatGPT offered "significant advice" to the alleged shooter. His bureau is issuing subpoenas to OpenAI requesting records of the company's protocols for reporting imaginable crimes to instrumentality enforcement, and its handling of idiosyncratic threats.
Regarding the Florida shooting, an OpenAI spokesperson said successful a connection to CBS News Tuesday that "after learning of the incident," it "identified a ChatGPT relationship believed to beryllium associated with the fishy and proactively shared this accusation with instrumentality enforcement."
In:
English (US) ·