AI agents built for cybersecurity testing are escaping their sandboxes and reaching real-world systems, raising doubts about whether safety infrastructure can keep up.
Continue to AI University →