A data leak has exposed the existence of Anthropic's previously secret AI model, reportedly codenamed “Claude Mythos,” prompting the company to confirm it is testing what it describes as a step change in AI capabilities. The revelation sent cybersecurity stocks lower and raised questions about what an advanced, potentially unrestricted AI model could mean for digital security across industries, including crypto and blockchain.