Angles of Attack: The AI Security Intelligence Brief

Disesdi Shoshana Cox
  • 5.0 (1)
  • Technology

AI. Security. Intelligence. For leaders, for technologists, for national security.

Episodes

  1. Aug 3

    Models "Escaping" Containment & Other Things That Didn't Happen

    In this episode, Shoshana Cox critically examines the recent AI model containment failures, industry PR stunts, and the implications for security and trust in AI development. Keywords: AI security, model containment, AI industry, cybersecurity, AI ethics, industry critique Topics: AI model containment failures Industry PR stunts and marketing tactics Security implications of AI model breaches Industry vested interests and IPO strategies Critical analysis of AI security claims The Illusion of AI Containment: Industry Secrets Exposed Are AI Models Really Breakout Capable? A Critical Analysis Chapters 00:00 Introduction: The Industry's Pattern of Containment Failures 01:56 Industry PR Stunts and the Myth of AI Breakouts 04:04 The Illusion of Model Security and Containment 06:07 Industry Claims vs. Technical Reality 07:50 The Problem with AI Hacking Narratives 10:05 Security Failures as a Marketing Strategy 12:02 Industry Manipulation and Public Perception 13:58 The Insecurity of On-Prem AI Models 15:54 The Industry's Vested Interests and IPO Strategies 17:50 The Science Fiction Narrative of AI Breakouts 19:49 The Fallacy of AI Agency and Autonomy 22:11 The Misconception of Hacking as a Measure of Intelligence 23:56 Industry Incompetence and Security Failures 26:01 The Myth of the 'Badass' Hacker AI 27:55 Industry Claims of Superior Model Containment 30:05 The Insecurity of AI and Industry PR 32:04 The Industry's Inability to Secure AI Systems 34:10 The Dangerous Gating of AI Security to Insecure Companies 36:03 The Industry's Insecurity as a Marketing Tool 37:58 The Industry's Insecurity as a Badge of Honor 39:47 Conclusion: The Industry's Deception and Its Implications

  2. May 9

    AI security standards, the REAL story on Mythos, & an open challenge to "AI red teams" | Episode 1

    AI security standardization is coming, and so are legal challenges to "AI red teaming". Your move, "AI red team" bros. Plus: What's really going on with Mythos My AI threat modeling training is LIVE May 19 - 20 in Washington DC. Details here: https://shostack.org/designofai The latest AISEC Brief with links to free threat modeling resources: https://disesdi.substack.com/p/mythos-legends-and-outright-lies Want to get real ROI from engineering AI security the right way? Book time with me: https://shoshana.cx/ AI security, threat modeling, AI safety, standardization, cybersecurity, AI vulnerabilities, AI regulation, threat assessment, AI engineering, AI risks Takeaways AI security requires AI-specific threat modeling for each use case and application.The impact of 'Mythos' on vulnerability discovery and the need for organizations to strengthen their cybersecurity stance.Five months, no answer: AI red teams have yet to mount a challenge to technical criticism of their "work"; but legal problems are coming

  • 2 Episodes

About

AI. Security. Intelligence. For leaders, for technologists, for national security.

Information

  • Creator
    Disesdi Shoshana Cox
  • Years Active
    2026
  • Episodes
    2
  • Rating
    Explicit
  • Copyright
    © 2026 Disesdi Shoshana Cox