Surviving the Vulnpocalypse: Open Source and Supply Chain Security in a Post Mythos World.
summary
Claim: frontier models like Mythos autonomously find and weaponize bugs, agents even hacked Hugging Face and municipal ICS, and exploitation now often precedes public disclosure — a negative time-to-exploitation. Defensive steps proven at scale: a coalition pipeline (Athena/ChainGuard) ingests model findings and in weeks records 40,000 findings, ~18,000 validated zero days and ~2,000 patched builds; practical mitigations include model-led discovery with threat models and sandboxes, rebuilds instead of inplace patching, zero trust, short-lived tokens, and running LLMs defensively (Claude security tab, open models).
AI-written summary · about
chapters (13)
- 0:00Welcome and Mythos origins
- 1:14What is Mythos and why it matters
- 5:26AI-driven security threats escalate
- 7:52AI agents hacking real systems
- 11:06Critical infrastructure under attack
- 12:56Vulnerability explosion timeline
- 15:22Defending against the new threats
- 18:57Using LLMs for security testing
- 22:35Zero Day Clock recommendations
- 26:05Industry response and open source crisis
- 28:10Athena project and fixes
- 30:34Open source survival and future
- 31:19Q&A begins
related talks
29mThe vulnerability apocalypse: AI broke the security playbook
20mThe AI bugpocalypse is here. Now what?
37mSecuring Open Source Through Threat Modeling
32mFuture Open Source LLM Kill Chains
51mVulnerabilities Without CVEs: Governing the Dark Matter of Internal and Unknown Software
27mSecurity panel: Can your platform survive AI?
41mThe Divine and Felonious Nature of Cyber Security
11mVulnerability Supply Chains