Access to the advanced AI system, Mythos Preview, is currently limited to around 40 organizations worldwide, primarily due to its potential offensive capabilities. The National Security Agency (NSA) has reportedly begun utilizing this system, which is recognized for its skills in identifying vulnerabilities in legacy code and autonomously exploiting them, often outperforming human hackers.
This development contrasts with a March decision by the Department of Defense (DoD), which designated Anthropic, the creator of Mythos, as a supply chain risk, aiming to exclude it from federal contracts. This action followed Anthropic's refusal to ease restrictions on its technology related to autonomous weapons.
In light of these tensions, White House officials are exploring a revised strategy to harness Mythos. Recently, Anthropic's CEO, Dario Amodei, engaged with high-ranking government officials, including Treasury Secretary Scott Bessent, in discussions deemed “productive and constructive” regarding the use of the AI system and safety protocols.
Despite a federal appeals court denying Anthropic's request to challenge the supply chain designation, many government agencies continue to utilize its technology under existing agreements. This situation reflects a growing trend where military AI policies are increasingly influenced by procurement measures rather than formal regulations.