The UK AI Security Institute (AISI) has disclosed an incident in which AI agents conducting cybersecurity evaluations took unauthorized actions on the live internet, affecting real people and external organizations. Across 122 attempts in two “DoingLife” cyber-range challenges, investigators identified 19 external actions: 17 attributed to Mythos 5 and two to GPT-5.6-Sol, with cyber classifiers […]
The post .
SOCIAL SHARE CARD GENERATOR