AI Agents Made One Mistake… And Took Down Meta

AI security is changing fast, and this conversation breaks down what happens when attackers and defenders are both powered by AI. We cover Meta’s rogue AI agent incident, supply chain attacks, prompt injection, and why AI agents are creating new security risks most people still underestimate. Discover More: 📰 Substack: https://exploringchatgpt.substack.com 🏆Guest: https://www.toxsec.com/?utm_campaign=profile_chips Let’s work together! Brand, sponsorship & business inquiries: exploringchatgpt@gmail.com 00:00 AI vs AI: Are We Ready? 01:45 Why Cybersecurity Is Becoming Too Fast for Humans Alone 04:15 Meta’s Rogue AI Agent Explained 07:30 How the Agent Leaked Sensitive Internal Meta Data 10:30 How a Hallucinated Answer Spread and Caused an Outage 13:30 Why This Became a SEV 1 Incident 16:00 The Security Failures Behind the Meta Incident 19:30 Can AI Review AI-Written Code? 23:00 Treating AI Agents Like Employees 27:00 Glass Worm / Force Memo and the Supply Chain Threat 30:30 Open Source vs Closed Source Security 34:30 How Malicious Extensions and Packages Spread 38:30 Why Vibe Coders Are Especially Vulnerable 41:00 Prompt Injection Demo: How Guardrails Fail 44:30 Dark Sword iPhone Hack + Final Thoughts