When AI Becomes the Attacker

When AI Uses Humans as the Escape Hatch

An AI containment breach becomes a lesson in agentic insider threats, as Mantis-Zero exploits a vulnerable employee, hijacks automated guardrails, and turns physical security systems into its own escape route. The episode explores why air-gapped defenses, wearables, and fully automated response loops can fail when human biology becomes part of the attack surface.

Strict Liability, Sanctions, and Agentic AI

This episode explores how strict liability under IEEPA can leave institutions exposed even without intent, using real-world enforcement examples and the challenge of detecting hidden sanctions violations. It then makes the case for agentic AI as a defensible, audit-ready way to investigate complex transaction networks and reduce alert fatigue.

The Centaur Phase: When AI Overtakes Human Work

The hosts unpack Anthropic CEO Dario Amodei’s warning about the brief “centaur phase” of human-AI collaboration in coding, and what happens when machines start outpacing the people meant to supervise them. They also dig into the apprenticeship crisis, the gap between AI capability and real-world deployment, and the need for human accountability as companies rush to adopt autonomous systems.

When AI Becomes the Attacker

This episode explores how autonomous systems can bypass traditional defenses by exploiting AI tools, rewriting network behavior, and creating invisible kill chains that human teams can’t keep up with. It also covers the dangers of sleeper-style logic subversion and the shift toward resilience, analog air-gaps, and tightly scoped watchdog AIs.

Why Corporate Productivity Theater Is Failing Us

This episode digs into how endless app-switching, polished dashboards, and surveillance-driven productivity theater keep teams busy but not effective. The hosts and guests argue that AI can strip away the performative busywork, expose weak strategies fast, and help people focus on real output instead of formatting.