Anthropic Plans Public Release of Its Powerful Mythos Model

Anthropic announced its intention to eventually release its powerful Mythos class bug-finding AI models to the general public once sufficient guardrails are developed. Currently, the technology remains restricted under Project Glasswing, an exclusive access program that was recently expanded to include US and allied governments. During its initial phase, Mythos scanned over 1,000 critical open source projects, uncovering 23,019 flaws, including a severe cryptographic bypass in the wolfSSL library. However, the sheer volume of high quality vulnerability reports has severely overwhelmed open source maintainers, exposing a massive structural bottleneck in human patching capacity.

Read the full story on The Register →