Anthropic has built an AI model that can autonomously find and exploit zero-day software vulnerabilities at a level the company says surpasses decades of human security research and every automated tool in existence. The model, Claude Mythos Preview, has uncovered long-hidden flaws in systems such as OpenBSD, FFmpeg, and core Linux components, and demonstrated it can rapidly turn known bugs into full working exploits at low cost.
Anthroplic reports that Mythos has found critical weaknesses in major cryptography libraries and protocols like TLS, AES-GCM, and SSH, raising urgent security concerns for DeFi and other crypto infrastructure that rely on friction-based defenses such as multisig, timelocks, and audits. The risk is particularly high for DeFi protocols, which are open source software where code is publicly readable by anyone, including a model like Mythos that can autonomously catalog every weakness in a codebase at machine speed.
While roughly $200 billion is locked in smart contracts across Ethereum, Solana, and other chains, the CoinDesk DeFi Select Index has gained 7% in 24 hours, outperforming bitcoin and ether. However, traders may want to keep an eye on developments around Mythos, given its potential implications for software and blockchain security. The model will not be released to the general public yet and is instead shared with select software giants under 'Project Glasswing.'
