Bad Claude, French Fridge Fiasco Fries Fromage, Teenage AI, SonicWALL, Aaran Leyland - SWN #632
Summary
Host Doug White discusses the security risks of AI agents, including a vulnerability in Anthropic’s document tools and the potential for sensitive data leakage via public GitHub repositories. The episode also covers a joint US government warning on Iranian cyberattacks targeting French refrigeration systems, a critical SonicWall SAML authentication flaw, and the impending obsolescence of 2G smart meters. Guest Aaron Leland explains how fake gift card scams exploit psychological biases to harvest verification codes and personal data.
Topics discussed
Intro and episode 632 overview
Risks of AI agents and autonomous network access
Sponsors: ThreatLocker and Government Security Summit
France blames Iran for cyberattacks on refrigeration systems
Joint US agency warning on refrigeration security
SonicWall SAML authentication bypass flaw
UK smart meters and the 2G network sunset risk
Claude 'Office Party' document processing vulnerability
OpenAI launches Teen Mode for ChatGPT
Interview with Aaron Leland on gift card and phone scams
Claude Code leaking sensitive data to public GitHub
Outro and subscription call to action
Listen ad-free on Castria