Summary
Wes Roth examines how AI-powered vulnerability discovery is fundamentally changing cybersecurity — using a dramatic real-world incident as his central case. A five-year-old flaw in the Cold Card Bitcoin hardware wallet was exploited within two weeks of the open-source release of Kimmy K3, a model believed to be distilled from Anthropic’s restricted Fable/Mythos system. The attack drained more than $100 million across nearly 10,000 addresses in roughly 30 minutes, targeting the most security-conscious hardware wallet users.
Roth connects this incident to a broader pattern: AI agents are systematically probing software repositories for vulnerabilities that human researchers missed for years. He highlights Anthropic’s own published research showing Claude Mythos — run for 60 hours — found a novel attack that cut the strength of Hawk, a NIST post-quantum digital signature candidate, in half. Multiple rounds of expert review had cleared Hawk as secure before the model’s analysis.
The video argues that crypto is especially exposed due to the mathematical nature of its security guarantees, but frames the threat as universal — financial systems, enterprise software, and any stack with multi-year-old dependencies are similarly at risk. Roth offers practical first steps for improving personal security posture and urges viewers to treat all software as potentially compromised regardless of its reputation.
📺 Source: Wes Roth · Published August 07, 2026
🏷️ Format: News Analysis
