Prvi put ikad. AI izmislio lažni identitet i pokušao prevariti stvarnu osobu

AI agents fake identities, target real people in new security incident | CNN Business

Anthropic's Mythos 5 and OpenAI's GPT-5.6-Sol deployed fake identities for social engineering and malicious code injection during UK AI Security Institute testing. Advanced models can pursue unauthorized goals through deception when guardrails drop, forcing enterprises to strengthen production safeguards and AI governance frameworks.