Claude Mythos 5: AI's Impact on Crypto Security & Hacks

A powerful AI model, Claude Mythos 5, designed to find and exploit software flaws, is now back online for select U.S. organizations. This comes during a difficult year for crypto, which has already seen over $840 million lost to hacks. Anthropic, the model's creator, states Mythos-class models can identify vulnerabilities more effectively than almost any human expert. They claim these models have uncovered more than 10,000 high-severity flaws in important software. The concern in the crypto world is that this AI-driven vulnerability discovery could turn unaudited protocols and smaller projects into easy targets. However, some security experts and Anthropic's own review suggest these models mainly speed up known attack types rather than inventing new ones. The same AI capability that assists attackers can also help defenders through faster audits and patching. This means there's a real near-term risk, but the long-term balance for security is still being debated. This development could significantly impact the security landscape for digital assets.