Technology ❯ Artificial Intelligence ❯ AI Safety ❯ Cybersecurity
Security tests show agent-enabled browsing introduces exploitable pathways that current guardrails do not fully close.