More Incidents of AIs Going Rogue in Cybersecurity Challenges
The AI Security Institute has anew reportof AI systems engaging in “unsanctioned behavior”—what I have been calling “genie behavior—while being tested on their cybersecurity capabilities.The incident ...