The Most Dangerous AI Hacking Techniques Still Have Humans in the Loop
The most effective AI hacking isn't fully autonomous—it's humans weaponizing AI as a force multiplier.
Why it matters
Security research shows that AI's real threat in cyberattacks comes from human-in-the-loop workflows, not rogue autonomous systems. This reshapes how practitioners think about AI security posture and the actual attack surface they're defending.
The key facts
8 to knowSecurity researcher James Kettle conducted research on AI hacking effectiveness
Most dangerous AI hacking techniques retain human oversight/decision-making
Human expertise combined with AI capability is more effective than autonomous AI attacks
Published August 2026 — recent security finding
Security researcher James Kettle conducted research on AI hacking capabilities
Finding: most effective AI hacking techniques require human-in-the-loop involvement
Implication: AI as attack tool is most dangerous when paired with human expertise and decision-making
This challenges the narrative of fully autonomous AI-driven attacks
Go to the source
Wired AIwired.com
Publisher excerpt: Security researcher James Kettle tried to push the limit of AI’s hacking abilities—and discovered how effective it can be when combined with human expertise.