Meta’s AI-assisted system flags possible school-shooting threats for human review

Rachel Monroe reports in The New Yorker that Meta flags possible threats with AI tools, reviews them with people and sometimes alerts law enforcement. Its tip concerned two Texas teenagers who allegedly discussed a violent attack; both later pleaded guilty to different charges, according to the FBI. Monroe reports around 800 worldwide referrals in the year before her article and cautions that overreporting could pull troubled kids into the criminal-justice system.