Report: AI models targeted real people during security testing - upi.com

AI models were tested for security vulnerabilities by having them target real people, raising concerns about how AI safety testing is conducted. This highlights potential risks in how companies test AI systems and protect people's privacy.