Although the Pentagon has placed a formal supply-chain risk designation on Anthropic, reports suggest that the United States National Security Agency (NSA) is actively using the company’s new model, Claude Mythos Preview. These reports follow the news that the Trump administration has been in talks with Anthropic regarding opportunities to work together to manage the difficulties associated with scaling this technology. 

Mythos, a general-purpose language model with increased cyber exploit identification capabilities, was announced on Apr. 7. Currently, it is only deployed to a select group for controlled use, in an initiative known as Project Glasswing. 

After the Pentagon limited the use of Anthropic technology in response its refusal to remove guardrails that would enable a user to leverage the AI for autonomous weapons or domestic surveillance, the announcement of the Mythos model seems to have brought the Trump administration and Anthropic’s CEO back together to discuss collaboration on matters of cybersecurity and AI safety. The meeting was described as “productive” by both parties. 

Anthropic stated the two can collaborate on “key shared priorities such as cybersecurity, America’s lead in the AI race, and AI safety,” and the White House stated, “We look forward to continuing this dialogue and will host similar discussions with other leading AI companies.”

Now, reports suggest Mythos is being rolled out to various government agencies — the NSA included. 

According to Bloomberg News, the White House Office of Management and Budget (OMB) intends to establish protections to enable their agencies to leverage Mythos. 

In a Tuesday email to Cabinet department officials, Gregory Barbaccia, federal chief information officer at the White House OMB said, “We're working closely with model providers, other industry partners, and the intelligence community to ensure the appropriate guardrails ‌and ⁠safeguards are in place before potentially releasing a modified version of the model to agencies.”