Mythos & Glasswing, What we actually know
summary
Anthropic’s Mythos demonstrates a genuine jump: it finds 271 vulnerabilities on Mozilla’s test bench, completed the long‑running cyber‑range challenge in ~17 of 100 attempts, and can decompile binaries and spin services up in containers to validate exploits; yet Project Glasswing finds only one low‑severity issue in curl. The real change is tempo — time from discovered bug to working exploit shrinks — so more actors can weaponize classic flaws (buffer overflows, injections, XSS) even if no novel vulnerability class appears.
AI-written summary · about
chapters (26)
- 0:00Mythos & Glasswing intro
- 1:04Mythos claims post-mythos era
- 2:33AI cybersecurity capabilities
- 3:56AI code analysis
- 5:21AI patching vulnerabilities
- 6:24Exploit creation significance
- 7:27Anthropic's claims trust
- 8:50Anthropic's tech docs
- 10:17Mythos binary reverse engineering
- 11:42Multi-threaded exploit thinking
- 13:04Live hacking CTF
- 14:12Mythos claims verification
- 14:34Mozilla Mythos vulns
- 17:23AI Security Institute tests
- 20:33Curl audit with Mythos
- 23:04Mythos threat level
- 24:29Vuln-to-exploit timelines
- 27:00Security vuln backlog
- 29:29Glasswing limits open source
- 31:55AI future capabilities
- 32:16Software security recs
- 33:21Vulnerability management steps
- 34:47Defensive security role
- 35:29Open source funding
- 36:54Conclusion call to action
- 37:16Q&A AI security
related talks
29mThe vulnerability apocalypse: AI broke the security playbook
36mSurviving the Vulnpocalypse: Open Source and Supply Chain Security in a Post Mythos World.
48mA Weakness In Chrome & A New Hacking Technique
39mThe AI Arms Race in Vulnerability Management, Who’s Winning?
35mStrangerDanger: Finding Security Vulnerabilities Before They Find You!
40mThe Mechanisms That Enable Misuse and What We Can Do About It
41mLive Hacking: Breaking into Your Web App
56mModern Web Vulnerabilities 2020