Anthropic advances Department of War initiative, outlines current status

Anthropic’s chief executive, Dario Amodei, detailed the narrow legal footing of the Department of War’s supply‑chain risk designation in a March 5 blog post, noting that the statute invoked—10 U.S.C. § 3252—“requires the Secretary of War to use the least restrictive means necessary to protect the supply chain.” According to the statement, the designation applies only to Claude deployments that are a direct component of Department‑of‑War contracts, not to any ancillary use by contractors’ other customers. Amodei emphasized that the law “exists to protect the government rather than to punish a supplier,” and that even for contractors the ruling “doesn’t (and can’t) limit uses of Claude or business relationships … if those are unrelated to their specific Department of War contracts.” The company therefore plans to continue providing Claude to warfighters and national‑security analysts at “nominal cost and with continuing support from our engineers” for any permitted use, while it prepares a legal challenge to the designation.

The administration’s stance has escalated beyond the statutory notice. Reuters reported that Defense Secretary Pete Hegseth summoned Amodei for “tough talks” over the military use of Claude, signaling a high‑level push to enforce the risk label. Wired added that President Donald Trump publicly announced a ban on Anthropic from all federal systems, a move that coincided with the Secretary of War’s X post declaring the supply‑chain risk and a contemporaneous Pentagon deal with OpenAI. Amodei’s blog acknowledges the “difficult day” these announcements created, and he apologized for an internal memo that was leaked to the press, clarifying that the memo was written six days earlier and does not reflect his current position.

Despite the political pressure, Anthropic maintains that its collaboration with the Department of War has yielded concrete operational tools. In the same March 5 statement, Amodei listed applications that have already been fielded: intelligence analysis, modeling and simulation, operational planning, and cyber‑operations support. He reiterated that Anthropic “has been very proud of the work we have done together with the Department,” but also stressed that the company “does not believe … that it is the role of Anthropic or any private company to be involved in operational decision‑making—that is the role of the military.” The firm’s only policy objections remain limited to “fully autonomous weapons and mass domestic surveillance,” which Amodei framed as high‑level usage concerns rather than day‑to‑day operational decisions.

Amodei’s post also highlighted ongoing negotiations aimed at preserving the narrow exceptions carved out by the law. He wrote that “we had been having productive conversations with the Department of War over the last several days, both about ways we could serve the Department that adhere to our two narrow exceptions, and ways for us to ensure a smooth transition if that is not possible.” The company is prepared to keep Claude available “for as long as we are permitted to do so,” while simultaneously assembling a legal team to contest the risk designation in court. The statement concludes with an apology for the tone of the leaked internal memo, noting that it “does not reflect my careful or considered views” and was “out‑of‑date” by the time of its publication.

The broader AI‑defense landscape is shifting rapidly. TechCrunch’s coverage of the same summons noted that Anthropic’s dispute arrives as the Pentagon finalizes a multi‑billion‑dollar partnership with OpenAI, effectively positioning a rival model as the default for many federal contracts. Analysts cited by Reuters and Wired have interpreted the Department of War’s move as an attempt to pressure Anthropic into aligning its policies with the administration’s more aggressive stance on AI governance, especially in the wake of the Trump administration’s push to “ban Anthropic from the US government.” Anthropic’s legal challenge, therefore, is not only about a single statutory interpretation but also about the future architecture of AI procurement across the defense establishment.