THE LONG VIEW

Bottom Line Up Front: The Pentagon’s demand for unrestricted access to Anthropic’s AI technology poses a significant threat to responsible AI governance, erodes trust with private innovators, and risks setting a precedent for coercive use of national security authority over ethical safeguards. Threat Identification: The U.S. Department of Defense, under Secretary Pete Hegseth, has issued an ultimatum to Anthropic—allow unrestricted military use of its AI model Claude by February 28, 2026, or lose its government contract and potentially be designated a supply chain risk [1]. This includes possible invocation of the Defense Production Act, a Cold War-era law enabling federal control over private industry [2]. Anthropic has resisted, citing its ethical commitment to prevent use in mass surveillance of Americans or fully autonomous weapons—uses it explicitly prohibits [3]. Probability Assessment: High probability (85%) that the Pentagon will follow through on contract termination if terms are not met by the February 28 deadline, based on Secretary Hegseth’s public ultimatum and pattern of centralizing military decision-making authority. Moderate probability (60%) that the Defense Production Act will be invoked, given its historical precedent but high political and legal stakes. Escalation to broader AI developer coercion is likely if unchallenged, with potential expansion to other AI firms by Q3 2026. Impact Analysis: The consequences are severe and multifaceted. First, undermining corporate-led AI safety standards risks normalizing unchecked military AI deployment, increasing potential for unlawful surveillance or autonomous systems without human oversight [4]. Second, public confrontation damages government-industry collaboration essential for national security innovation. Third, designating ethical AI firms as 'supply chain risks' could deter future private investment in defense-aligned AI, weakening U.S. technological edge. Finally, bipartisan congressional alarm—voiced by Senators Tillis and Warner—signals growing institutional concern over governance failure [5]. Recommended Actions: 1) Immediately pause the ultimatum and move negotiations into classified, closed-door channels to preserve strategic partnerships. 2) Establish an independent AI Oversight Task Force under Congress to define lawful and ethical boundaries for military AI use. 3) Enact binding AI governance legislation for defense applications to prevent ad hoc coercion. 4) Clarify publicly that mass surveillance and lethal autonomous weapons remain prohibited under current law and policy. Confidence Matrix: - Threat Identification: High confidence (supported by direct statements from Amodei, Parnell, and Hegseth) - Probability Assessment: Medium-high confidence (informed by stated deadlines and historical DoD actions, but dependent on internal decision-making not fully public) - Impact Analysis: High confidence (based on documented ethical policies, legal frameworks, and bipartisan reactions) - Recommended Actions: Medium confidence (feasible policy responses, though political will remains uncertain) Citations: [1] O’Brien, M., & Toropin, K. (2026, February 26). US Warns Anthropic to Allow Unrestricted Use of AI by Military. Associated Press via Bloomberg.com. [2] Id. [3] Amodei, D. (2026, February 26). Statement on Pentagon Contract Negotiations. Anthropic. [4] Parnell, S. (2026, February 26). Social media statement. X (formerly Twitter). [5] Warner, M. (2026, February 26). Statement on Pentagon-Anthropic Dispute. U.S. Senate Intelligence Committee. —Marcus Ashworth