AI gets hacked, and BitLocker gets bypassed
Summary
Graham Cluley and Paul Ducklin discuss how AI coding agents can be hijacked via fake Sentry error reports to execute malicious code, bypassing traditional security. They also cover Nightmare Eclipse’s full disclosure of zero-day exploits that bypass Microsoft Defender and BitLocker, sparking a debate on responsible disclosure. The episode concludes with a look at the FBI’s kinetic cyber range, a simulated town used for training law enforcement in cyber investigations.
Topics discussed
Intro: AI hacks, fake breaches, and episode overview
Sponsor: ProtonPass password manager features
AI Coding Agents: The Sentry error injection attack
Analysis: Why the Sentry attack works and vendor response
Sponsor: Vanta compliance automation
Nightmare Eclipse: Microsoft bugs and BitLocker flaws
Sponsor: CoreView Microsoft 365 security checks
Pick of the Week: FBI cyber range and Linux hotspot
Guest: Sun on AI agent security risks and permissions
Outro: ProtonPass trial, credits, and ZipRecruiter ad
Listen ad-free on Castria