Smashing Security podcast #471: This AI worm just rewrote its own rules

This Smashing Security podcast episode reviews multiple security stories: researchers at the University of Toronto built and tested an AI-driven adaptive worm capable of discovering exploits, spreading autonomously, and recruiting infected machines' compute for distributed AI tasks; Meta’s automated customer-support AI exposed a flaw enabling attackers to change account emails and seize high-profile Instagram accounts; AI tools have rapidly surfaced many new FFmpeg vulnerabilities; and separate stories include a leaked METV awards access key incident and a hijacked shipment theft. The discussion highlights the ease with which freely available models and modest coding can be combined to create powerful offensive capabilities and the resulting challenges for defenders.