White House Moves To Give Us Agencies Access To Anthropic'S Mythos

The White House is pushing to grant US government agencies access to Anthropic's Mythos Preview, an AI model that has been withheld from broad public release because of its offensive cybersecurity capabilities, according to Bloomberg.

During testing, Mythos reportedly identified thousands of high-severity vulnerabilities, including security flaws that had gone undetected for more than two decades. Anthropic has so far distributed the model only to a curated group of technology and cybersecurity firms under a programme called Project Glasswing.

The move comes shortly after Treasury Chief Information Officer Sam Corcos sought access to the model for vulnerability-hunting purposes, and following a meeting convened by Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell with bank chief executives to discuss emerging AI-related threats to the financial system.

The push also comes against a complicated legal backdrop. The Pentagon designated Anthropic a supply chain risk earlier this year, and a federal appeals court recently rejected the company's bid to block that designation. A separate ruling out of San Francisco, however, allows Anthropic to continue working with non-Department of Defense agencies in the meantime.