The Artificial Intelligence Strategy for the Department of War is the standing AI policy of the US defense establishment, issued not as a statute or executive order but as a memorandum from Secretary of War Pete Hegseth dated 9 January 2026 and published three days later. It directs the Department to become an "AI-first" warfighting force, sets seven named programs with six-month demonstration deadlines, and makes two procurement changes that have since governed the Department's relationships with frontier developers: model-objectivity benchmarks as a primary procurement criterion, and standard "any lawful use" language in every contract through which AI services are procured. The signed text is summarized at Artificial Intelligence Strategy for the Department of War (Hegseth memorandum, 2026).
Status and authority
The strategy is an internal policy instrument. It binds the components of the Department — the military departments, combatant commands, defense agencies and field activities — and creates no obligation on any private party except through the contract terms it directs. It invokes Executive Order 14179 and America's AI Action Plan as its policy authorities. It does not amend DoD Directive 3000.09, the Department's standing directive on autonomy in weapon systems, and does not mention it.
Key provisions
Programs. Seven Pace-Setting Projects run in fiscal year 2026 across warfighting (Swarm Forge, Agent Network, Ender's Foundry), intelligence (Open Arsenal, Project Grant) and enterprise (GenAI.mil, Enterprise Agents) mission areas, each with a single accountable leader, monthly demonstration to the Deputy Secretary and the Under Secretary for Research and Engineering, and initial demonstration by transition-partner users within six months. Every component must name at least three fast-follower projects within 30 days.
Model availability. GenAI.mil is directed to put "America's world-leading AI models directly in the hands of our three million civilian and military personnel, at all classification levels."
Procurement cadence. The Chief Digital and AI Office must establish a vendor cadence enabling "the latest models to be deployed within 30 days of public release," which the memorandum makes "a primary procurement criterion for future model acquisition."
Data access. Components must maintain federated data catalogs at all classification levels and deliver current catalogs within 30 days; the CDAO is authorized to direct release of any departmental data to cleared users with a valid purpose, and denials must be justified within seven days to the Under Secretary for Research and Engineering.
Barrier removal. A monthly Barrier Removal Board under the Under Secretary for Research and Engineering may waive non-statutory requirements. Exercises that do not meaningfully incorporate AI and autonomous capabilities are subject to review by the Director of Cost Assessment and Program Evaluation "for resourcing adjustment."
Architecture. Program managers acquiring AI capabilities must enforce Modular Open System Architectures and expose interfaces sufficient for third-party integration without prime-contractor support.
Model conduct. Under the heading "Clarifying 'Responsible AI' at the DoW — Out with Utopian Idealism, In with Hard-Nosed Realism," the strategy directs that the Department "must not employ AI models which incorporate ideological 'tuning' that interferes with their ability to provide objectively truthful responses to user prompts" and "must also utilize models free from usage policy constraints that may limit lawful military applications." The CDAO was given 90 days to establish benchmarks for model objectivity as a primary procurement criterion; the Under Secretary for Acquisition and Sustainment was given 180 days to incorporate standard "any lawful use" language into any contract procuring AI services.
Consequences
The "any lawful use" directive is the provision from which the Department's dispute with Anthropic followed. Anthropic's usage policy prohibited autonomous weapons targeting and mass surveillance of Americans; the Department treated those restrictions as the "usage policy constraints" the strategy bars, and the dispute proceeded to litigation, where the summary-judgment opinion quotes the strategy's Responsible AI subsection (Anthropic PBC v. U.S. Department of War — Order on Cross Motions for Summary Judgment (N.D. Cal., August 27, 2026)). OpenAI accepted equivalent terms in March 2026 and Google in April 2026, the latter on "any lawful government purpose" language, making three frontier-lab outcomes under one contract clause (Source: theinformation.com).
The GenAI.mil directive to place models in the hands of all personnel at all classification levels was cited by CNN on 18 September 2026 in its account of a false AI-assisted intelligence report that nearly led to the boarding of a Chinese vessel (Source: cnn.com).
Sens. Mark Warner, Christopher Coons and Jack Reed, in a letter to Hegseth and Director of National Intelligence Jay Clayton dated September 19, 2026, cited the strategy as one that "treats AI governance policies as 'blockers' to AI adoption in military contexts" and asked for Inspector General investigations of the Minab strike and the SOCPAC false-intelligence episode. The strategy's section "Wartime Approach to Blockers" names "blockers to data sharing, Authorizations to Operate (ATOs), test and evaluation and certification, contracting, hiring and talent management, and other policies that inhibit rapid experimentation and fielding"; the phrase "AI governance policies" is the letter's characterization (Sens. Warner, Coons and Reed — letter to Secretary Hegseth and DNI Clayton on AI targeting and intelligence errors (2026); Artificial Intelligence Strategy for the Department of War (Hegseth memorandum, 2026)).
Dating
Reporting has placed the originating memorandum on 12 January 2026, the date it was posted. The document itself is dated 9 January 2026, the date used in the summary-judgment opinion. Both the "any lawful use" directive and the phrase "the risks of not moving fast enough outweigh the risks of imperfect alignment," attributed in reporting to a 12 January memo, appear in the 9 January text, which indicates one instrument rather than two (Artificial Intelligence Strategy for the Department of War (Hegseth memorandum, 2026)).
Relationships
- depends-on: Artificial Intelligence Strategy for the Department of War (Hegseth memorandum, 2026) — the signed text.
- depends-on: Executive Order 14179 — Removing Barriers to American Leadership in Artificial Intelligence, America's AI Action Plan — the invoked policy authorities.
- related: DoD Directive 3000.09 — Autonomy in Weapon Systems — the autonomy directive the strategy leaves untouched.
- deployed-by: DOD — Department of Defense (AI Deployer) — the implementing department.
- contradicts: Sens. Warner, Coons and Reed — letter to Secretary Hegseth and DNI Clayton on AI targeting and intelligence errors (2026) — Senate letter criticizing the strategy's treatment of "blockers."
- related: Pete Hegseth — the signatory.
- supports: Anthropic v. United States (Pentagon ban challenge) — supplies the contract clause at issue.
- related: Autonomous Weapons — the capability areas the Pace-Setting Projects address.