Trump signs a memorandum placing the 'most advanced AI' under military control and prohibiting vendors from shutting it down.

      TL;DR: Trump signed NSPM-11, mandating swift military AI integration and prohibiting vendors from disabling models without permissions. Hegseth is tasked with updating the autonomous weapons policy.

      On Friday, President Trump signed a national security presidential memorandum requiring the U.S. military and intelligence agencies to expedite the implementation of advanced AI technologies. This directive, NSPM-11, sets a framework for the quick integration of top-tier AI models from various vendors and forbids any company from disabling, degrading, or altering an AI system essential to warfighters without prior government consent.

      This vendor restriction is particularly notable, as it prevents an AI firm from withdrawing a deployed model from military operations on its own, even if it has safety concerns regarding its use. This clause relates to the ongoing dispute between the Pentagon and Anthropic, which was categorized as a supply chain risk after it declined to permit its Claude models for use in autonomous weaponry or mass surveillance.

      “The men and women who defend our nation deserve the best, most secure and most reliable AI in the world,” stated Michael Kratsios, director of the White House Office of Science and Technology Policy.

      The memorandum also instructs Defence Secretary Pete Hegseth to release an updated directive concerning autonomous weapon systems within 90 days. This revision will amend DoD Directive 3000.09, which outlines the Pentagon's policy on the use of autonomous and semi-autonomous weapons, including the necessity for human judgment prior to the application of lethal force.

      There are certain limits defined in the memo. It forbids defense agencies from developing or releasing AI models intended to “censor free speech, embed ideological bias or conduct unlawful surveillance against the American people.” However, it does not clarify these terms or describe how compliance would be monitored.

      This directive comes on the heels of Tuesday’s executive order, which established a voluntary 30-day review period for advanced AI models before they are made public. Together, these documents illustrate a dual strategy: light regulation for the commercial sector alongside an assertive military adoption.

      The mention of “multiple vendors” indicates a shift from reliance on a single provider. Previously, Anthropic was the sole approved AI vendor for classified military applications. Following classified agreements with Nvidia, Microsoft, and AWS, the administration is now formalizing a multi-vendor strategy.

      The memo emphasizes accountability, requiring commanders, directors, and agency heads to ensure AI is utilized in accordance with its specified obligations. Annual assessments of key guidance within the national security sector are mandated to stay aligned with AI advancements. It remains uncertain whether these reviews will be substantive or merely symbolic.