This AI worm just rewrote its own rules
Summary
Graham Cluley and James Ball discuss a new AI worm that autonomously rewrites its own rules to bypass security controls, highlighting the risks of adaptive malware. They also examine a vulnerability in Meta’s customer service AI, which allowed attackers to hijack accounts by tricking the bot into resetting passwords to unauthorized emails. The episode touches on James Ball’s PhD research into how legal systems view AI in surveillance versus copyright cases.
Topics discussed
Intro: James Ball's PhD on AI and Law
Sponsor: Opswat and proactive security
AI Worm: Self-replicating malware research
AI Worm: Bypassing controls and finding zero-days
Sponsor: Vanta automates compliance
Meta AI: Instagram account takeover vulnerability
Sponsor: Expo autonomous pentesting
Pick of the Week: Final Fantasy VII Remake
Outro: Credits and listener shoutouts
Listen ad-free on Castria