Mythos vs. Reality: Zero-Day Defense for 2026 Supply Chains

Frontier‑scale AI is rapidly redefining how organizations confront zero‑day threats buried deep within software and firmware supply chains. Anthropic’s Project Glasswing showed that models like Claude Mythos Preview can uncover severe vulnerabilities at speeds traditional analysis simply cannot match. Yet despite the industry’s fixation on AI‑enhanced attackers, the more grounded reality is that even the most advanced systems remain only as secure as the vendors and contractors supporting them. Recent unauthorized access attempts involving Mythos‑class environments made this clear: the failure point wasn’t AI containment but basic supply‑chain weakness.

As AI‑assisted analysis grows more capable, it is exposing systemic flaws across cloud, software, and firmware ecosystems that organizations have long struggled to detect. This session explores why SBOM‑centric strategies, though useful, cannot keep pace with machine‑speed discovery and why zero‑day exposure increasingly originates in third‑party and firmware components. The widening gap between traditional assurance methods and AI‑driven analysis demands a new approach to validation and monitoring.

Attendees will learn how to operationalize AI for continuous validation, strengthen vendor assurance programs, and harden architectures against rapidly evolving threats. By understanding where current approaches fall short and how frontier‑scale models reshape defensive strategy, organizations can build supply‑chain security capable of withstanding AI‑accelerated risk.

Speakers:

• Colin Whittaker (Moderator)
• Erik Thoen, VP of Product Management, ReversingLabs
• Janet Worthington, Senior analyst at Forrester
• Paul Asadoorian, Principal Security Researcher at Eclypsium & founder of Security Weekly
• John Sapp, Field CISO, Chainguard