Risky Business
Risky Business

Nov 18, 2026 · 1h 4m

AI agents accelerate cyberattacks without replacing human judgment

Risky Business #860 -- Google: Agentic hacking has “crossed a threshold”

The episode examines how quickly offensive AI is scaling, where human control still matters, and whether security policy can keep pace.

3 key takeaways
  1. 1AI agents can automate credential discovery and exploitation while humans still choose targets and direct operations.
  2. 2Offensive agents are becoming more parallelized and stealthy, increasing the importance of constrained deployment and identity testing.
  3. 3A sprawling security agenda links AI breaches to fragile infrastructure, supply-chain attacks, surveillance policy, and export controls.

Don't miss

James Wilson recounts Artex persistently attempting to bypass a CAPTCHA and then spray credentials, illustrating the practical behavior of offensive agents.

The brief

Patrick Gray, James Wilson, and Jason Kikta frame agentic hacking as a threshold event: AI can accelerate discovery and exploitation, but humans still select targets and steer many operations.

The panel moves from an Optimizely supply-chain compromise and Java and PHP flaws to aging infrastructure, arguing that deferred maintenance often makes major outages predictable.

James Wilson describes Artex repeatedly trying to bypass a CAPTCHA before spraying credentials, offering a concrete test of how much effort offensive agents can sustain.

Jason Kikta expects offensive agents to become more parallelized and stealthy, while warning that defenders must constrain scope and test identity systems carefully.

The episode closes by connecting AI-assisted security operations with Fireworks AI breaches, export controls, Section 702, drone supply chains, and formal verification.

Weekly email

State of the Pod, every Monday

The week's most-discussed names and stories from podcasts, with the moments behind them. Free, five minutes, unsubscribe in one click.

Listen to the full episode and explore every guest, topic, and moment on PodLume.

AI agents accelerate cyberattacks without replacing human judgment · PodLume