Anthropic intends to eventually release models matching its Mythos bug-finding AI to the general public, once it can build adequate safety measures, the company stated in an update on its restricted-access program, per The Register. Mythos, revealed in April, is skilled enough at spotting security flaws in code that Anthropic limited it to select participants under “Project Glasswing” out of concern that open access would let cybercriminals find and exploit vulnerabilities first. The company plans to widen access to more partners, including U.S. and allied governments, before any broad release, and acknowledged that no company has yet developed safeguards strong enough to fully prevent misuse. Mythos has scanned over 1,000 open-source projects and surfaced an estimated 6,202 high or critical vulnerabilities, including a flaw in the wolfSSL cryptography library used by billions of devices. Some maintainers have asked Anthropic to slow its disclosures, citing limited capacity to patch.
