Anthropic’s Mythos model found vulnerabilities in classified US government systems – report (opens original article in a new tab)

Anthropic's Mythos model found vulnerabilities in US government systems during a test, but did not exploit them, leading to export restrictions by the US government.

• Anthropic's Mythos model identified vulnerabilities in US government systems during testing
• Mythos did not exploit the vulnerabilities within the time it took to identify them
• US government restricted exports of Mythos and Fable models citing national security concerns