Pentagon Chief Technology Officer: Anthropic still on the blacklist, Mythos exception handling

The Pentagon technology chief said to Congress on May 1, “Anthropic is still on the blacklist, but Mythos Preview is another matter,” formally acknowledging that the U.S. Department of Defense is treating Anthropic’s main brand and its new model, Mythos, differently. The statement aligns with Axios’ April 19 disclosure that the NSA (National Security Agency) has already put Mythos into use, and it also officially sets the policy direction of a “Mythos exception pipeline”—while at the same time suggesting an internal contradiction in legal arguments, since the Pentagon had argued in court that using Anthropic tools would threaten national security, which clashes with current practical actions.

Anthropic vs Pentagon standoff in March: the dispute over refusal to authorize behind Hegseth’s blacklist

The conflict between Anthropic and the Pentagon stems from February 2026. Defense Secretary Pete Hegseth required Anthropic to agree to the deployment of its models for “autonomous lethal attacks” and “mass domestic surveillance of U.S. citizens.” CEO Dario Amodei refused. After Hegseth then gave Amodei a deadline but still received no concession, Trump directly ordered all federal agencies to “immediately stop” using Anthropic technology. In March, Hegseth formally listed Anthropic as a “supply chain risk” company, and in April Anthropic appealed to the Second Circuit Court of Appeals seeking a temporary restraining order, which failed to materialize.

However, even if the official stance is a blacklist, actual usage by federal agencies has been inconsistent. Axios’ April 19 exclusive report said the NSA has deployed Mythos Preview—Anthropic’s most powerful model, with access limited to 40 agencies by Anthropic itself, described officially as “possessing excessively dangerous network attack capabilities.” In other words, Mythos is “dangerous enough to have access restricted,” yet it is also “dangerous enough that federal agencies must use it.”

Mythos exception pipeline: Amodei meeting Wiles and Bessent on 4/30

On April 30, Amodei met with White House chief of staff Susie Wiles and Treasury Secretary Scott Bessent, explicitly discussing the conditions under which Mythos is used within the government and Anthropic’s security practices. Around the same time, Anthropic advanced a new round of fundraising with an estimated valuation of $900B. The high-level meeting between Wiles + Bessent was seen as a signal that the Trump administration is reassessing its Anthropic policy. The next step leaked after the meeting was “how agencies other than the Pentagon” will adopt Mythos—namely, bypassing the Pentagon blacklist and having other federal agencies procure directly.

The May 1 statement by the Pentagon technology chief—“Mythos is another matter”—is precisely the legal packaging for this strategy. In effect, the Pentagon splits the handling of “no procurement of Anthropic’s main brand” (maintaining alignment with Hegseth’s position) and “Mythos can be an exception” (echoing the White House and the NSA’s substantive needs), creating “two tracks” in legal argumentation. But the feasibility of this split depends on whether Mythos can be sufficiently separated from Anthropic’s main entity at the contract level, potentially involving a separate subsidiary or a separate authorization structure.

What to watch next: whether Hegseth endorses it, and how much strategy space is opened for other frontier model vendors

The next key point is Hegseth’s official response to the Pentagon technology chief’s statement. If Hegseth agrees to a “Mythos exception,” it would mean that the Anthropic blacklist since February is effectively relaxed. If Hegseth maintains the original position, policy division would emerge within the Pentagon and could trigger direct White House intervention to arbitrate. From an industry perspective, the outcome of this case will affect the strategic options for other frontier model developers such as OpenAI, Google DeepMind, and xAI when faced with military requests for “authorization of autonomous lethal attacks.” If Anthropic refuses but can still preserve certain federal procurement (through the Mythos exception), it would provide other companies with a concrete precedent for maintaining commercial relationships while not granting full authorization.

This article, Pentagon technology chief: Anthropic still blacklisted, handling Mythos as an exception, first appeared on Chain News ABMedia.