Cybersecurity looks like proof of work now

Anthropic released Mythos, a specialized security LLM kept from public release but provided to critical software makers for vulnerability discovery. Independent validation from the AI Security Institute confirms Mythos substantially outperforms previous models on complex network attack simulation benchmarks. The analysis frames security as a token-cost equation: systems are hardened by spending more tokens on vulnerability discovery than attackers will spend exploiting them.