Policy
Pentagon Tests OpenAI and Google Models as Anthropic Faces Supply-Chain Designation
The U.S. Department of Defense is evaluating alternative AI models from OpenAI and Google as it seeks to reduce reliance on Anthropic, which faces a controversial supply-chain risk designation.
By Michael C ·
The U.S. Department of Defense is actively evaluating alternative AI models from OpenAI and Google as it seeks to reduce its reliance on Anthropic, according to a senior defense official. The testing initiative, which began in March 2026, involves 25 Pentagon 'power users'—senior officials and military strategists—evaluating competing models to determine which best serves the Department's operational needs.
The move follows Defense Secretary Pete Hegseth's controversial decision in February to designate Anthropic as a 'supply-chain risk' over the company's insistence on safety guardrails for its technology. Anthropic has since filed suit challenging the designation, arguing that the decision could cost the company billions in revenue and sets a dangerous precedent for government interference in AI development.
The Supply-Chain Risk Designation
Hegseth's supply-chain risk determination represents an unprecedented intervention by the U.S. government into the AI market. The designation, made under authorities typically reserved for foreign companies or suppliers with ties to adversarial nations, effectively bars Anthropic from certain Pentagon contracts and raises questions about the company's eligibility for future defense work. The move signals a fundamental shift in how the U.S. government views AI companies—no longer as neutral technology providers, but as strategic assets whose values and priorities matter to national security.
Anthropic's crime, according to Pentagon officials, was its refusal to remove safety constraints from its AI models. The company maintains that its guardrails—designed to prevent misuse of AI for harmful purposes—are essential to responsible AI deployment. Defense officials counter that these constraints limit the models' utility for military applications and create an unacceptable dependency on a company that may prioritize ethics over national security. The tension reflects a deeper disagreement about whether AI safety measures are features or bugs in military contexts.
The Competitive Evaluation
The Pentagon's testing of OpenAI and Google models reflects the geopolitical stakes of AI development. Both companies have been more accommodating of military applications, though both have also maintained some ethical guidelines. OpenAI's GPT-5 and Google's Gemini Omni represent the current frontier of large language model capability, and both have demonstrated effectiveness in tasks ranging from strategic analysis to logistics optimization. The evaluation process itself is politically charged, as it signals that the Pentagon is willing to shop around for AI providers willing to meet its specifications.
Broader Implications for AI Governance
The Anthropic dispute raises fundamental questions about the role of government in directing AI development. Should the state be able to compel private companies to remove safety features from their products? Or do companies have the right to maintain ethical standards even if it costs them government contracts? These questions pit national security concerns against corporate values and individual conscience.
These questions will likely shape AI policy for years to come. If the Pentagon succeeds in forcing Anthropic to choose between its values and its revenue, other AI companies may face similar pressure to compromise on safety and ethics. Conversely, if Anthropic prevails in its legal challenge, it may establish important precedent protecting AI companies' ability to maintain safety standards. The international dimension adds another layer of complexity. China and other competitors are watching closely to see whether the U.S. government will fracture its AI industry through heavy-handed intervention. A weakened Anthropic benefits China's AI development efforts by reducing competition and potentially driving talent and resources to less ethically constrained competitors.