Anthropic experiences second major security breach in days, exposing hundreds of thousands of lines of code for its AI coding tool and revealing details about the 'Mythos' model in development.

Hundreds of thousands of lines of source code were accidentally leaked, providing external researchers with access to proprietary information

The breach exposed internal architecture details and insights into Anthropic's upcoming AI models, including one codenamed 'Mythos'

This marks the second significant data exposure for Anthropic within days, raising concerns about the company's security practices

The leaked code gives competitors and researchers visibility into Anthropic's AI development roadmap and technical infrastructure