Pentagon Scrutinizes Anthropic's AI Guardrails for Military Use
The Pentagon is reportedly threatening to revoke contracts with AI vendor Anthropic if the company does not loosen its model guardrails for military applications. This development underscores the growing tension between commercial AI safety standards and government defense requirements. It also signals increasing cross-border regulatory divergence on the acceptable uses of advanced AI.
- The central conflict involves Anthropic's refusal to remove "red lines" that prohibit its AI from being used for mass domestic surveillance of Americans and for fully autonomous weapons that can operate without human oversight. The Pentagon is demanding use of the technology for "all lawful purposes," arguing that legal compliance is its responsibility as the end-user, not the vendor's. - During a meeting, Defense Secretary Pete Hegseth gave Anthropic CEO Dario Amodei a deadline of Friday, Feb. 28, to agree to the military's terms. The Pentagon has threatened to cancel the $200 million contract, invoke the Defense Production Act to compel compliance, or designate Anthropic a "supply chain risk," which would effectively blacklist it from government-related work. - Anthropic is currently the only AI company with a model deployed on the Pentagon's classified networks. However, other major AI labs, including OpenAI, Google, and xAI, also received similar $200 million contracts, with OpenAI and xAI reportedly agreeing to the Pentagon's broader terms. - The Department of Defense has a publicly stated strategy to "lead in military ethics and AI safety," having adopted five ethical principles for AI in 2020 which include ensuring systems are "Responsible" and "Governable." - Coinciding with the dispute, Anthropic updated its Responsible Scaling Policy (RSP), removing a prior commitment to halt the development of models it could not prove were safe. The company cited the need to remain competitive as the reason for the change, stating it did not make sense to make "unilateral commitments" if competitors were advancing. - This standoff highlights a significant shift from 2018, when widespread employee protests led Google to withdraw from the Pentagon's Project Maven. More recently, both Google and OpenAI have removed previous prohibitions on military applications from their usage policies. - The disagreement occurs in a global regulatory vacuum, as there is no international treaty governing the military use of AI. The European Union's AI Act, for example, explicitly excludes military applications from its scope, showcasing the fragmented approach to AI governance worldwide.
