The DoD AI ethical principles, explained

The DoD AI Ethical Principles are five commitments the Department of Defense (DoD) adopted for its use of artificial intelligence (AI): responsible, equitable, traceable, reliable, and governable. The Department's release of February 24, 2020 says they apply to both combat and non-combat functions and build on the military's existing ethics framework, which the release grounds in the Constitution, Title 10 of the U.S. Code, the law of war, and existing treaties. Later documents date the adopting Secretary of Defense memorandum to February 21, 2020, three days before the release. A 2021 memo and a 2022 pathway set out how to implement them, and a 2023 directive on autonomous weapons cites them.

The five principles, in the Department's words

The release says the Department adopted the principles following recommendations the Defense Innovation Board made the previous October, after 15 months of consultation with experts and the public. The release gives the principles in this wording:

• Responsible. DoD personnel will exercise appropriate levels of judgment and care, while remaining responsible for the development, deployment, and use of AI capabilities.
• Equitable. The Department will take deliberate steps to minimize unintended bias in AI capabilities.
• Traceable. The Department's AI capabilities will be developed and deployed such that relevant personnel possess an appropriate understanding of the technology, development processes, and operational methods applicable to AI capabilities, including with transparent and auditable methodologies, data sources, and design procedure and documentation.
• Reliable. The Department's AI capabilities will have explicit, well-defined uses, and the safety, security, and effectiveness of such capabilities will be subject to testing and assurance within those defined uses across their entire life-cycles.
• Governable. The Department will design and engineer AI capabilities to fulfill their intended functions while possessing the ability to detect and avoid unintended consequences, and the ability to disengage or deactivate deployed systems that demonstrate unintended behavior.

The release named the Joint Artificial Intelligence Center (JAIC) as the implementation focal point. The 2021 memo restates the principles with small wording differences. The memo's Traceable drops "the" before technology, and the memo's Reliable ends with testing across AI capabilities' entire life-cycle.

From principles to practice, 2021 and 2022

Deputy Secretary of Defense Kathleen Hicks issued the memo Implementing Responsible Artificial Intelligence in the Department of Defense on May 26, 2021. The memo reaffirms the principles and applies them to all DoD AI capabilities of any scale, including AI-enabled autonomous systems, for warfighting and business uses. The memo sets six foundational tenets for responsible AI (RAI): RAI Governance, Warfighter Trust, AI Product and Acquisition Lifecycle, Requirements Validation, Responsible AI Ecosystem, and AI Workforce. Under Warfighter Trust, the memo calls for a test and evaluation framework that integrates real-time monitoring, algorithm confidence metrics, and user feedback. It gave the RAI Working Council 150 days to deliver a final pathway.

The Responsible AI Strategy and Implementation Pathway, dated June 2022, is that document. The pathway names trust as the desired end state. For each tenet it lists lines of effort, with goals, offices of primary responsibility, and estimated time horizons where known. The goal for warfighter trust is for system operators to reach justified confidence in AI-enabled systems. The pathway's toolkit lists AI data card and model card templates, and directs a test and evaluation master plan template for AI and standard solicitation language for AI capabilities.

DoDD 3000.09, a separate policy on weapons

DoD Directive (DoDD) 3000.09, Autonomy in Weapon Systems, took effect January 25, 2023. The directive replaces the November 21, 2012 version and was approved by Deputy Secretary Hicks. It governs autonomous and semi-autonomous functions in weapon systems. It does not apply to cyberspace capabilities, unarmed platforms, unguided or manually guided munitions, mines, unexploded ordnance, or autonomous systems that are not weapon systems.

The directive defines an autonomous weapon system as one that, once activated, can select and engage targets without further intervention by an operator. Paragraph 1.2.a requires that such systems be designed so that commanders and operators can exercise appropriate levels of human judgment over the use of force. Unless a system falls within listed exceptions, the Under Secretaries of Defense for Policy and for Research and Engineering and the Vice Chairman of the Joint Chiefs of Staff must approve it before formal development. Approval is required again before fielding, with the Under Secretary of Defense for Acquisition and Sustainment in place of Research and Engineering.

The directive links the two policies in paragraph 1.2.f: AI in such weapons must be consistent with the ethical principles and the 2022 pathway. As of October 21, 2026, the Directives Division list still shows the January 25, 2023 version.

Changes in 2025 and 2026

Executive Order 14347 of September 5, 2025 allows the Department to be called the Department of War as a secondary title. Statutory references to the Department of Defense remain controlling until law changes them. The 2020 release now sits on war.gov.

On January 9, 2026, the Secretary of War issued the memo Artificial Intelligence Strategy for the Department of War. A section headed "Clarifying 'Responsible AI' at the DoW" says the Department must not use models with ideological tuning that interferes with objectively truthful responses, and must use models free from usage policy constraints that may limit lawful military applications. The memo directs the Chief Digital and AI Office (CDAO) to set model-objectivity benchmarks as a primary procurement criterion within 90 days. The Under Secretary of War for Acquisition and Sustainment is to add standard "any lawful use" language to AI service contracts within 180 days. The CDAO is to ensure all existing AI policy guidance aligns with the memo. The memo does not name the 2020 principles or DoDD 3000.09.

What to check

• Which text of the principles does a document quote: the February 24, 2020 release or the May 26, 2021 memo?
• Does a requirement trace to a principle, a tenet, or a line of effort in the June 2022 pathway?
• For a weapon system, does DoDD 3000.09 apply or does an exception cover it, and are the senior approvals on file?
• Is a "reliable" claim backed by testing within explicit, well-defined uses across the life cycle?
• Has a later document been aligned with the January 9, 2026 memo, and does a contract carry its "any lawful use" language?

Azcuna's ELLMENT, the E-2D Large Language Model Entity, was proposed to NAVAIR PMA-231 under Navy SBIR Phase I topic DON26BZ01-NV010. The prototype, built at private expense on publicly available doctrine, answers with a page-level citation and refuses when the documents do not support an answer. The proposal has been submitted.

Sources

1. 1DOD Adopts Ethical Principles for Artificial Intelligence · U.S. Department of Defense release (February 24, 2020), now hosted on war.gov
2. 2Implementing Responsible Artificial Intelligence in the Department of Defense · Deputy Secretary of Defense memorandum (May 26, 2021)
3. 3U.S. Department of Defense Responsible Artificial Intelligence Strategy and Implementation Pathway · DoD Responsible AI Working Council (June 2022)
4. 4DoD Directive 3000.09, Autonomy in Weapon Systems · Office of the Under Secretary of Defense for Policy (effective January 25, 2023)
5. 5DoD Directives · Executive Services Directorate, Directives Division
6. 6Artificial Intelligence Strategy for the Department of War · Secretary of War memorandum (January 9, 2026)
7. 7Executive Order 14347, Restoring the United States Department of War · Federal Register (September 10, 2025)

Where this shows up in our work

• CALIBER, the market testbed for AI agents measures from an AI agent's trades whether it is gullible, herds, overpays, or can be steered.
• ELLMENT, the offline LLM for E-2D crews answers a crew's question from the aircraft's own manuals and cites the page, with no network.