Anthropic embedded engineers at the NSA while fighting the Pentagon ban in court

The National Security Agency is using Anthropic's most restricted AI model — and Anthropic has sent engineers there to help — despite an active Pentagon order banning federal agencies from working with the company. 1

In February 2026, Defense Secretary Pete Hegseth designated Anthropic a supply-chain risk to national security after the company refused to let its models be used for mass domestic surveillance or fully autonomous weapons. President Trump posted on Truth Social directing all federal agencies to cease use of Anthropic products, with a six-month phase-out window. 2 Anthropic said it would challenge the designation in court.

The dispute followed a contract renegotiation where the Pentagon demanded access to Claude for "all lawful purposes." Anthropic drew two hard lines — no mass domestic surveillance, no autonomous weapons. The Defense Department walked.

On June 5, the Financial Times reported that Anthropic has embedded around half a dozen engineers inside the NSA to help the agency deploy Mythos for cyber operations. 1 TechCrunch confirmed the reporting. The NSA declined to confirm or deny it; Anthropic did not respond to a request for comment.

This is the clearest picture yet of how the NSA is using the model. An April Axios scoop had already established that the NSA was accessing Mythos despite the Pentagon ban — one source told Axios the NSA was among the unnamed organizations Anthropic granted access to when it quietly restricted the model to roughly 40 entities. 3 The June 5 report goes further: Anthropic is not just tolerating the NSA's use, it is actively staffing it.

What the engineers are specifically doing at the NSA remains unclear. TechCrunch noted it's uncertain whether Mythos is being used in active hacking operations at this point. The NSA's core offensive role is conducting cyberattacks on foreign adversaries — the same class of task that Mythos Preview demonstrated a capability for when Anthropic's red team found it could discover thousands of high-severity zero-day vulnerabilities. 4

The NSA sits under the Department of Defense organizationally, which means Hegseth's ban technically applies to it. But intelligence agencies operate with significant operational independence, and the government's demand for frontier cyber AI has not gone away because a contracting dispute soured relations.

The April Axios report put it plainly: the military is now broadening its use of Anthropic's tools while simultaneously arguing in court that using those tools threatens U.S. national security. 3 Some officials want the fight to end so they can use what Anthropic is building. Others still argue Anthropic's refusal to comply proved it cannot be trusted.

Dario Amodei met White House chief of staff Susie Wiles and Treasury Secretary Scott Bessent in April to discuss Mythos access within government, and both sides described the meeting as productive — a signal that the ban may be closer to a negotiating position than a final outcome. 3

Anthropic is preparing for an October IPO. The company filed a confidential S-1 on June 1 and has named Goldman Sachs, Morgan Stanley, and JPMorgan Chase as lead underwriters. 5 The NSA story lands in the middle of that roadshow preparation.

For investors reading the prospectus, the Pentagon dispute is a disclosed risk — but the FT's reporting shows the actual exposure is more tangled. Anthropic is not simply locked out of government work; it is embedded inside the most operationally significant intelligence agency in the country while simultaneously fighting a lawsuit over its right to exist there. That is a better commercial position than the headline ban suggests, but it also means Anthropic's government revenue is running on informal agreements rather than contracts, which creates its own IPO disclosure complications.

The UK's AI Security Institute has also disclosed access to Mythos through its own channel. 3 Anthropic restricted the model to roughly 40 organizations globally, citing the risk that its offensive cyber capabilities could be exploited at scale. The NSA is apparently one of the 40.