In February, the U.S. Department of Defense halted a contract after conflicts over the use of AI, but the situation changed after Mithos was unveiled on the 7th of an undisclosed month. Anthropic is now in discussions with the White House and other organizations.
Reports have emerged that the National Security Agency (NSA) is using the security-focused AI model “Claude Mithos Preview” developed by Anthropic, which is on a U.S. government blacklist. Despite being sanctioned by its superior institution, the Department of Defense, the NSA has been reportedly utilizing Mithos.
While specific details of the NSA’s use of Mithos are unclear, Axios noted that other organizations have used it to identify security vulnerabilities in their systems. Initially, Anthropic restricted access to Mithos to about 40 organizations due to its powerful ability to identify security vulnerabilities and even generate attack codes, raising concerns about potential misuse.
Anthropic revealed only 12 of these organizations, and according to sources, the NSA was one of the undisclosed entities granted access. This situation is notable as it follows a major conflict between the Department of Defense and Anthropic earlier in the year over clauses restricting large-scale surveillance and the use of autonomous weapons.
Donald Trump, during his presidency, decided to expel Anthropic from federal government dealings, and the Department of Defense designated the company as a “supply chain risk company,” the first U.S. company to receive such a designation. However, the dynamics shifted after Anthropic released Mithos on the 7th of an undisclosed month.
The Department of Defense has shown a dual approach by arguing in a “supply chain risk company (blacklist) invalidation lawsuit” that Anthropic’s AI model threatens national security while simultaneously discussing its use.
Axios assessed that the Trump administration’s focus on strengthening cybersecurity seems to take precedence over the conflict between the Department of Defense and Anthropic. Anthropic’s CEO, Dario Amodei, met with White House Chief of Staff Suzy Whiles and Treasury Secretary Scott Besant on the 17th of an undisclosed month to discuss the administration’s potential uses for Mithos and Anthropic’s future plans and security practices.
According to sources, both parties considered the discussions productive, with a focus on how departments other than the Department of Defense might utilize Mithos. However, Anthropic, the Department of Defense, NSA, and the Office of the Director of National Intelligence did not respond to Axios’s request for confirmation on the matter.