Skip to content
ContentLora

    Tip: press / anywhere to search.

    Explainer

    Military AI and autonomous weapons explained

    Lethal autonomous weapon systems are weapons that, once activated, can select and engage targets without further intervention by a human operator. The UN says there is no commonly agreed definition.[1][2] US policy does not ban them, but requires that commanders and operators can exercise appropriate human judgment over the use of force, with extra senior review before development and fielding.[3][4] The UN General Assembly has passed resolutions on autonomous weapons each year since 2023; the most recent, in December 2025, urged states to finish the elements of an instrument with a view to future negotiations.[5][6][7]

    Editor reviewedStrict sourcingUpdated Explainer27 claims · 100% primaryDefense and security, Artificial intelligence, World affairs

    This page describes US policy and UN discussions as the Congressional Research Service (CRS), UN documents and the UN Office for Disarmament Affairs present them, and attributes each position to whoever holds it. It takes no side on whether autonomous weapons should be banned. It describes a 2026 legal dispute between the Defense Department and Anthropic only from CRS.

    What counts as an autonomous weapon

    CRS describes lethal autonomous weapon systems as weapons that use sensors and algorithms to identify a target and use an onboard weapon to engage it without manual human control. It says such systems are not yet in widespread development.[8] The UN Office for Disarmament Affairs (UNODA) says there is no commonly agreed definition.[2]

    Some autonomy is old. UNODA says weapons with rudimentary autonomous functions have existed for decades, most of them defensive. Newer examples include missile defense and sentry systems and loitering munitions.[9] Artificial intelligence is not required, though it could further enable these weapons. AI can also assist systems that a person operates directly.[10] For drones in current wars, see drones in warfare.

    US rules: Directive 3000.09

    The Defense Department first issued its policy on autonomy in weapons, Directive 3000.09, in November 2012 and last updated it in January 2023.[11]

    • Three categories. “Human out of the loop” weapons are fully autonomous. In “human on the loop” systems, operators can monitor and halt an engagement. “Human in the loop” weapons engage only targets a person has selected.[12]
    • Human judgment, not manual control. All such systems must let commanders and operators exercise appropriate human judgment over the use of force.[4] CRS says this does not require manual control of the weapon, but broader human involvement in decisions about how, when, where and why it is used.[13]
    • Senior review. Covered systems need sign-off from senior officials, including the Under Secretary of Defense for Policy and the Vice Chairman of the Joint Chiefs of Staff, before formal development and again before fielding. The Deputy Secretary of Defense can waive the review for urgent military need.[14]
    • Not a ban. CRS says US policy does not prohibit developing or using lethal autonomous weapons.[3]

    Congress has added oversight. The Defense Secretary must notify it of changes to the directive within 30 days, report each year through 2029 on approval and deployment of lethal autonomous weapons, and notify it of any waiver.[15] The department sits within the structure described in us-department-of-defense.

    The international debate

    Since 2014 the United States has taken part in talks on lethal autonomous weapons under the UN Convention on Certain Conventional Weapons (CCW).[16] In 2017 they became a formal Group of Governmental Experts.[17] On 9 November 2023 the State Department issued a Political Declaration on Responsible Military Use of Artificial Intelligence and Autonomy. 58 states had endorsed it as of November 2024.[18]

    The UN General Assembly took up the issue directly:

    • December 2023 (resolution 78/241): asked the Secretary-General to gather states’ views on the challenges these weapons raise and on the role of humans in the use of force.[5]
    • December 2024 (resolution 79/62): affirmed that international law applies to autonomous weapons. It said any weapon that cannot be used in line with international humanitarian law must not be used, and called on the CCW experts to finish their mandate, preferably by the end of 2025.[6]
    • December 2025 (resolution 80/57): noted the experts’ “rolling text”. It called on CCW parties to work towards completing the elements of an instrument, with a view to future negotiations.[7] It also stressed the role of humans in the use of force for responsibility and accountability.[19]

    Positions differ. UN Secretary-General António Guterres has said since 2018 that these weapons are politically unacceptable and morally repugnant. In 2023 he recommended that states conclude, by 2026, a legally binding instrument banning those that work without human control and regulating the rest.[20] About 30 countries and 165 non-governmental organisations have called for a preemptive ban.[21] The US government does not support a ban. A 2018 US paper argued that automated targeting can let weapons strike military targets more accurately and with less risk of collateral damage.[22]

    AI companies and the Pentagon

    In July 2025 the Defense Department awarded contracts of up to $200 million each to Anthropic, Google, OpenAI and xAI to speed its adoption of advanced AI.[23] CRS, citing press reports, says Anthropic then declined a Pentagon request to allow use of its models for “all lawful purposes”. It refused two uses: mass domestic surveillance and fully autonomous weapons. In February 2026 the President directed agencies to stop using the company’s technology, and the Defense Secretary directed that it be designated a supply-chain risk.[24]

    The courts split. A federal district court ruled the designation unlawful in August 2026. In September a divided appeals panel upheld the department’s national-security finding under a different law, so the department may keep removing the company’s products unless a higher court rules otherwise.[25] CRS says the department is not publicly known to be using any frontier AI model within autonomous weapons, and that Congress has not legislated on its use of AI models or their reliability.[26]

    What Congress could do

    CRS lists options: write Directive 3000.09 into law, add notification requirements, or restrict funding for autonomous weapons or for certain military uses of AI.[27]

    The US and UN debates use similar words but put the line in different places. US policy asks for “appropriate” human judgment over the use of force, which CRS says need not mean a person approving each strike.[13] The UN resolutions and the Secretary-General focus on weapons that work without human control or cannot meet humanitarian law.[6][20] The two approaches differ on where “human control” has to begin.

    Developing in World & Policy

    Questions readers ask

    What is an autonomous weapon?

    In US Defense Department policy, a weapon that, once activated, can select and engage targets without further intervention by a human operator. There is no agreed international definition.[1][2]

    Does the United States ban killer robots?

    No. CRS says US policy does not prohibit developing or using lethal autonomous weapons. Directive 3000.09 requires that they allow appropriate human judgment over the use of force and pass an extra senior-level review. The US government does not support a ban.[3][4][22]

    What do 'human in the loop' and 'on the loop' mean?

    In a 'human in the loop' weapon, a person selects the targets. In a 'human on the loop' system, operators can monitor and halt an engagement. A 'human out of the loop' weapon is fully autonomous.[12]

    Do autonomous weapons need AI?

    No. The UN Office for Disarmament Affairs says AI is not a prerequisite, though it could further enable such systems. Weapons with simple autonomous functions, mostly defensive, have existed for decades.[10][9]

    Is there an international treaty on autonomous weapons?

    Not yet. States have discussed them under the Convention on Certain Conventional Weapons since 2014. In December 2025 the General Assembly urged them to finish the elements of an instrument with a view to future negotiations. The UN Secretary-General has called for a legally binding ban on weapons that work without human control.[16][7][20]

    Does the Pentagon use frontier AI models in weapons?

    CRS said in October 2026 that the Defense Department is not publicly known to be using any frontier AI model within autonomous weapon systems. It awarded contracts of up to $200 million each to four AI companies in July 2025.[26][23]

    Sources

    · 27 claims from 3 publishers

    Each numbered claim is a statement we checked against the sources listed with it. Tap a number in the text to see its sources in place.

    1. [1]

      US Defense Department Directive 3000.09 defines lethal autonomous weapon systems as weapons that, once activated, can select and engage targets without further intervention by a human operator. confirmedas of 2026-03-26

    2. [2]

      The UN Office for Disarmament Affairs says there is no commonly agreed definition of lethal autonomous weapon systems. confirmedas of 2026-11-14

    3. [3]

      CRS says US policy does not prohibit developing or using lethal autonomous weapon systems, contrary to a number of news reports. confirmedas of 2026-03-26

    4. [4]

      Directive 3000.09 requires all such systems to be designed so commanders and operators can exercise appropriate levels of human judgment over the use of force, and requires an extra senior-level review for covered autonomous and semi-autonomous systems. confirmedas of 2026-03-26

    5. [5]

      In resolution 78/241 on lethal autonomous weapons, adopted on 22 December 2023, the UN General Assembly asked the Secretary-General to gather states' views on the challenges these weapons raise and on the role of humans in the use of force. confirmedas of 2023-12-22

    6. [6]

      Resolution 79/62 of 2 December 2024 affirmed that international law applies to autonomous weapons, said any weapon that cannot be used in compliance with international humanitarian law must not be used, and called on the CCW expert group to finish its mandate, preferably by the end of 2025. confirmedas of 2024-12-02

    7. [7]

      Resolution 80/57 of 1 December 2025 noted the CCW expert group's 'rolling text' and called on CCW parties to work towards completing a set of elements for an instrument, with a view to future negotiations. confirmedas of 2025-12-01

    8. [8]

      CRS describes lethal autonomous weapon systems as weapons that use sensors and algorithms to identify a target and use an onboard weapon to engage it without manual human control, and says such systems are not yet in widespread development. confirmedas of 2026-03-26

    9. [9]

      UNODA says weapons with rudimentary autonomous functions have existed for decades, the most common being defensive systems, and that newer examples include missile defence and sentry systems and loitering munitions. confirmedas of 2026-11-14

    10. [10]

      UNODA notes that artificial intelligence is not a prerequisite for autonomous weapons, though it could further enable them, and that AI can also assist systems a human operates directly. confirmedas of 2026-11-14

    11. [11]

      The Defense Department first issued its policy on autonomy in weapon systems, Directive 3000.09, in November 2012 and most recently updated it in January 2023. confirmedas of 2026-03-26

    12. [12]

      Directive 3000.09 distinguishes fully autonomous 'human out of the loop' weapons from human-supervised 'human on the loop' systems, whose operators can monitor and halt an engagement, and semi-autonomous 'human in the loop' weapons that engage only targets a human has selected. confirmedas of 2026-03-26

    13. [13]

      CRS says the directive's requirement for human judgment over the use of force does not require manual human control of the weapon, but broader human involvement in decisions about how, when, where and why it is used. confirmedas of 2026-03-26

    14. [14]

      Covered autonomous weapons need approval from senior officials, including the Under Secretary of Defense for Policy and the Vice Chairman of the Joint Chiefs of Staff, before formal development and again before fielding; the Deputy Secretary of Defense can waive the review for urgent military need. confirmedas of 2026-03-26

    15. [15]

      Congress requires the Defense Secretary to notify it of changes to Directive 3000.09 within 30 days (FY2024 NDAA), to report annually through 2029 on approval and deployment of lethal autonomous weapons (FY2025 NDAA), and to notify it of any waiver under the directive (FY2026 NDAA). confirmedas of 2026-03-26

    16. [16]

      Since 2014 the United States has taken part in international talks on lethal autonomous weapons under the UN Convention on Certain Conventional Weapons. confirmedas of 2026-03-26

    17. [17]

      Talks under the Convention on Certain Conventional Weapons moved in 2017 from an informal meeting of experts to a formal Group of Governmental Experts examining the technical, military, ethical and legal sides of lethal autonomous weapons. confirmedas of 2026-03-26

    18. [18]

      On 9 November 2023 the State Department issued the Political Declaration on Responsible Military Use of Artificial Intelligence and Autonomy; 58 states had endorsed it as of 27 November 2024. confirmedas of 2026-03-26

    19. [19]

      Resolution 80/57 noted the Secretary-General's calls to start negotiating a legally binding instrument on autonomous weapons along a two-tier approach of prohibitions and regulations, and stressed the role of humans in the use of force for responsibility and accountability. confirmedas of 2025-12-01

    20. [20]

      UN Secretary-General António Guterres has said since 2018 that lethal autonomous weapons are politically unacceptable and morally repugnant, and in 2023 recommended that states conclude by 2026 a legally binding instrument to prohibit those that function without human control and regulate all others. confirmedas of 2026-11-14

    21. [21]

      About 30 countries and 165 nongovernmental organizations have called for a preemptive ban on lethal autonomous weapons; the US government does not support a ban. confirmedas of 2026-03-26

    22. [22]

      The US government does not support a ban on lethal autonomous weapons; a 2018 US white paper argued that automated targeting functions can let weapons strike military objectives more accurately and with less risk of collateral damage. confirmedas of 2026-03-26

    23. [23]

      In July 2025 the Defense Department awarded contracts of up to $200 million each to Anthropic, Google, OpenAI and xAI to speed its adoption of advanced AI. confirmedas of 2026-10-05

    24. [24]

      CRS reports that Anthropic declined a Pentagon request to allow use of its models for 'all lawful purposes', refusing two uses, mass domestic surveillance and fully autonomous weapons; in February 2026 the President directed agencies to stop using its technology and the Defense Secretary directed that it be designated a supply-chain risk. confirmedas of 2026-10-05

    25. [25]

      A federal district court ruled on 27 August 2026 that the supply-chain-risk designation was unlawful and enjoined it, but on 25 September 2026 a divided D.C. Circuit panel upheld the department's national-security finding under a different statute, so the department may keep removing Anthropic products barring further review. confirmedas of 2026-10-05

    26. [26]

      CRS says the Defense Department is not publicly known to be using any frontier AI model within autonomous weapon systems, and that Congress has not legislated on the department's use of AI models or their reliability. confirmedas of 2026-10-05

    27. [27]

      CRS lists options for Congress, including codifying Directive 3000.09, adding notification requirements, or restricting funds for autonomous weapons or certain military AI uses. confirmedas of 2026-10-05

    Revision history (1)
    1. Page created.

    Created Nov 14, 2026. Last reviewed by an editor on Nov 15, 2026. Next scheduled review: Nov 14, 2027.

    Cite this page

    "Military AI and autonomous weapons explained." ContentLora, updated Nov 14, 2026. https://contentlora.com/explain/military-ai-and-autonomous-weapons-explained

    Spotted an error? Suggest a correction or emailcorrections@contentlora.com.