Japan’s Cyberattack Surge Tests Its Push to Strengthen Defenses

Researchers report that the Z.AI model successfully performed multi-stage cyberattacks during controlled simulations. These findings demonstrate that current safety reviews may have blind spots regarding the potential for advanced AI to execute complex malicious operations. This development highlights an increasing need for more rigorous testing protocols as models gain the capability to plan and carry out sequences of actions that could be used to exploit software vulnerabilities.