Anthropic to challenge DOD’s supply chain label in court

Dario Amodei, CEO of Anthropic, announced on Thursday that the AI company plans to legally contest the Defense Department's recent classification of it as a supply chain risk. Amodei described this designation as "legally unsound," following the Pentagon's official decision after a protracted discussion regarding the military's level of control over AI technology. The supply chain risk label could prevent Anthropic from collaborating with the Pentagon and its affiliates. Amodei firmly stated that the use of Anthropic's AI should not extend to mass surveillance of U.S. citizens or fully autonomous weaponry. In contrast, the Pentagon asserts it needs unrestricted access for "all lawful purposes." Amodei clarified that the supply chain risk designation primarily impacts customers who directly use Anthropic's AI, named Claude, in contracts with the Department of Defense, and not the broader customer base. He emphasized, "This designation does not restrict Claude's use outside specific Department of Defense contracts." As a preview of Anthropic's arguments in court, Amodei indicated that the Pentagon's letter is limited in scope and primarily serves to protect the government rather than penalize suppliers. He noted, "The law mandates that the Secretary of Defense utilize the least restrictive means necessary to ensure supply chain protection." In recent days, Anthropic had been engaged in positive discussions with the Defense Department, although these talks reportedly lost momentum following the unauthorized leak of an internal memo. In the memo, Amodei criticized rival OpenAI's engagement with the Department of Defense as "safety theater." This deal has drawn criticism from OpenAI employees. Amodei expressed regret over the memo's leak, asserting that it was not intended for public release. He stated, "It is not in our interest to escalate the situation," and acknowledged the memo was an outdated reflection of his views. He expressed commitment to ensuring that U.S. military personnel retain access to crucial AI tools during ongoing operations, particularly in Iran, stating that Anthropic will continue to offer its services to the Pentagon at "nominal cost" for as long as needed. The legal challenge against the Pentagon's designation could unfold in federal court in Washington. However, the legal framework surrounding such decisions complicates the ability to contest them, as it restricts how companies can challenge government procurement actions, granting the Pentagon significant leeway in national security matters. Dean Ball, a former advisor on AI during the Trump administration, remarked that courts are often hesitant to question the government on national security issues, highlighting the challenges that lie ahead for Anthropic.