Anthropic Claude Mythos AI Sparks DeFi Exploit Debate After $73M Losses

The story has shifted from confirmed protocol breaches to a broader debate on AI-driven smart contract attacks. On June 10, Anthropic’s Claude Mythos model (Fable 5) triggered industry warnings that advanced AI could reduce exploit costs and skill barriers to near zero, especially for unaudited contracts. This follows June 9 reports of a $36 million Humanity Protocol loss from malware-compromised private keys and Chainalysis data showing $36.7 million in DeFi hacks over six months, dominated by the $26.2 million Truebit exploit. Analysts now emphasize that both operational security gaps and emerging AI tools are amplifying risks across the ecosystem.