One of the most contentious agencies in the US is reportedly sampling the highly potent Mythos AI developed by Anthropic.

      The agency's use of Mythos underscores an increasing divide within the US government regarding AI risks.

      The struggle over AI policy within the US government has grown more complex. Reports suggest that the National Security Agency (NSA) is utilizing Anthropic's Mythos Preview, despite senior Pentagon officials advocating for a ban on the company due to supply chain issues. This situation illustrates how urgent security needs can quickly surpass official directives.

      Since February, the Department of Defense has sought to restrict Anthropic and encourage other vendors to follow suit. However, as per an Axios report, the NSA seems to be proceeding with one of the company’s most robust models, indicating that cybersecurity demands are taking precedence over the current internal disputes within the government.

      Why access to Mythos is restricted

      Mythos is notable because Anthropic has maintained tight control over it. Reports indicate the company has limited access to approximately 40 organizations due to the model's offensive cyber functionalities, and only a handful of these users have been publicly acknowledged. One source indicated that the NSA is among those unnamed agencies with access.

      This scenario appears less like a typical chatbot rollout and more like a high-stakes security asset. Sources mentioned that those with access primarily use Mythos to scan their own systems for potential vulnerabilities, explaining why national security officials would still value it in light of the existing trust issues.

      Washington's AI paradox

      The major concern lies in the visible contradiction within the government. While one segment regards Anthropic as a risk, another is reportedly testing its leading model. This suggests that the blacklist is not as settled as previously claimed.

      The discord seems to extend beyond procurement matters. Defense officials wished for Anthropic to make Claude accessible for all lawful functions, whereas the company resisted applications related to extensive domestic surveillance and autonomous weaponry. Some officials interpreted this as evidence that Anthropic could not be relied upon by the military, a point the company disputes.

      What's next

      The forthcoming question is whether Mythos will remain an NSA-specific exception or pave the way for broader government access. Sources indicated that a recent discussion between Anthropic CEO Dario Amodei, White House chief of staff Susie Wiles, and Treasury Secretary Scott Bessent centered on Mythos's role in government and the company's overall security stance. Both parties characterized the meeting as productive.

      If more agencies adopt Mythos, this scenario may serve as a preview of how Washington navigates the challenges of powerful AI amid internal policy disputes concerning tools that officials are reluctant to relinquish.