Autonomous weapons

Autonomous weapons, usually discussed under the label lethal autonomous weapon systems (LAWS), are weapon systems that, once activated, can select and engage targets without further intervention by a human operator. The phrase covers a spectrum rather than a single machine: at one end sit shipboard guns that automatically detect, track and engage incoming missiles, and at the other sit hypothetical systems that would roam a battlespace choosing human targets on their own. The category is defined by where a human sits in the targeting decision, not by how sophisticated the underlying artificial intelligence happens to be ^[1][2].

There is no internationally agreed definition. States told the UN Secretary-General in 2024 that no such definition exists, and several argued that agreement on one is not a precondition for negotiating rules ^[3]. The United States uses a definition set out in Department of Defense Directive 3000.09; the International Committee of the Red Cross (ICRC) uses a functional one built around target profiles; and the UN body negotiating in Geneva has assembled its own working characterization in a chair's rolling text ^[1][4][5]. The disagreement is not merely semantic, because a definition that captures anti-radar loitering munitions and shipboard point defense implies that autonomous weapons already exist, while a definition built around machines that independently choose human targets implies they do not.

As of mid-2026 no treaty specifically regulates autonomous weapons. The Group of Governmental Experts (GGE) established under the Convention on Certain Conventional Weapons (CCW) has provisionally agreed 22 formulations toward a possible instrument, the UN General Assembly has adopted three resolutions on the subject, and the Secretary-General has repeatedly asked states to conclude a legally binding instrument by 2026 ^{[3][5][6][7][8]}.

Definitions and the human control spectrum

A widely cited operational definition comes from the US Department of Defense. DoD Directive 3000.09, reissued on 25 January 2023, defines an autonomous weapon system as "a weapon system that, once activated, can select and engage targets without further intervention by an operator," and adds that this includes operator-supervised systems that a human can override but that will otherwise keep engaging on their own ^[1]. The same glossary defines a semi-autonomous weapon system as one that "is intended to only engage individual targets or specific target groups that have been selected by an operator," which is where fire-and-forget homing munitions fall ^[1].

Advocacy and legal writing usually maps these categories onto the loop taxonomy set out in the 2012 Human Rights Watch report Losing Humanity: human-in-the-loop robots select targets and deliver force only with a human command, human-on-the-loop robots do so under the oversight of a human operator who can override, and human-out-of-the-loop robots select targets and deliver force without any human input or interaction ^[9]. The US Congressional Research Service maps DoD's three categories onto exactly this vocabulary ^[2].

Scroll sideways for more →

The ICRC's 2021 position frames the issue functionally: after activation such a weapon "self-initiates or triggers a strike in response to information from the environment," so the user does not choose, and may not even know, the specific target ^[4]. The phrase "meaningful human control" was advanced by the UK NGO Article 36 in an April 2013 policy paper, which argued for a positive obligation in international law for individual attacks to be under meaningful human control, noting that a human setting some parameters of a mission "alone is clearly not sufficient human control" ^[10]. The United States has consistently preferred "appropriate levels of human judgment over the use of force," a formulation CRS stresses does not require manual control of the weapon ^[1][2].

Systems already in service

Several fielded systems meet the DoD definition of autonomy in at least some modes, which is why claims that autonomous weapons do not yet exist need qualifying. Close-in defensive guns are the oldest example. The Phalanx Close-In Weapon System is described by its manufacturer as a rapid-fire, computer-controlled, radar-guided gun that carries out search, detection, threat evaluation, tracking, engagement and kill assessment, functions normally spread across several systems, firing between 3,000 and 4,500 rounds per minute against anti-ship missiles and other close-in threats ^[11]. It is installed on all US Navy surface combatant ship classes and on those of 24 allied nations ^[11]. DoD Directive 3000.09 exempts systems of this kind from its senior review requirement: operator-supervised autonomous systems used against materiel targets for local defense against time-critical or saturation attacks on installations or platforms with personnel aboard do not need the special approval ^[1].

Loitering munitions are the second category, and the more contested one. Israel Aerospace Industries describes its HARPY as "fully autonomous," equipped with an anti-radiation seeker that autonomously seeks and strikes emitting high-value targets, launched "without prior intelligence on the target's location" and able to loiter for up to nine hours, including in GNSS-denied environments ^[12]. Its electro-optical sibling HAROP is presented differently: IAI states that "as a guided weapon system, HAROP is supervised by remote human-in-the-loop mission control," and that a strike may be aborted if required ^[13]. The pair illustrates how narrow the technical gap between a supervised and an unsupervised munition can be.

Libya 2020 and the Kargu-2 report

The most frequently cited alleged instance of autonomous targeting comes from a March 2021 report by the UN Panel of Experts on Libya. Describing the 2020 offensive by Government of National Accord Affiliated Forces, the Panel wrote that retreating opposing units and logistics convoys "were subsequently hunted down and remotely engaged by the unmanned combat aerial vehicles or the lethal autonomous weapons systems such as the STM Kargu-2 and other loitering munitions," and that these systems "were programmed to attack targets without requiring data connectivity between the operator and the munition: in effect, a true 'fire, forget and find' capability" ^[14]. The Panel separately listed a Turkish transfer of the STM Kargu-2 loitering munition, dated to its identification on 27 May 2020, in its table of arms embargo transfer violations ^[14].

The report is routinely summarized in the press as documenting the first autonomous drone killing, and it does not say that. It attributes the engagements to unmanned combat aerial vehicles or to autonomous systems without separating the two. The following paragraph records that Pantsir S-1 air defence systems harassed by that combination "suffered significant casualties," but the Panel nowhere states that a weapon operating autonomously killed a particular person ^[14]. Zachary Kallenborn, writing in the Bulletin of the Atomic Scientists on 20 May 2021, put the ambiguity plainly: if anyone was killed in an autonomous attack it would likely represent a historic first, and the UN report heavily implies that they were but "is not explicit on the matter" ^[15]. The episode is a useful caution about how quickly a hedged UN finding hardens into a settled fact.

The ban campaign

Organized opposition predates the Geneva talks. Human Rights Watch and Harvard Law School's International Human Rights Clinic published Losing Humanity in November 2012, recommending that states prohibit the development, production and use of fully autonomous weapons through an international legally binding instrument ^[9]. The Campaign to Stop Killer Robots was formed in October 2012 and publicly launched in 2013; Human Rights Watch coordinated it from 2012 to 2021, and its steering committee includes Amnesty International, Article 36 and the International Committee for Robot Arms Control ^[36]. The coalition now reports more than 300 member organizations working across more than 75 countries, and campaigns for new international law on autonomy in weapons systems, framed as a demand for meaningful human control over the use of force ^[36].

The CCW process in Geneva

States began discussing autonomous weapons informally under the CCW in 2014 ^[2]. The Fifth Review Conference of the CCW, held in Geneva from 12 to 16 December 2016, adopted Decision 1 establishing an open-ended Group of Governmental Experts on emerging technologies in the area of LAWS, which first met in 2017 ^[16]. At their meeting of 13 to 15 November 2019, the High Contracting Parties endorsed 11 guiding principles affirmed by the GGE ^[17]. Those principles state that international humanitarian law continues to apply fully to all weapon systems; that human responsibility for decisions on the use of weapons systems must be retained because accountability cannot be transferred to machines; that weapons must be operated within a responsible chain of human command and control; and, more contentiously for campaigners, that such systems "should not be anthropomorphized" and that the CCW is an appropriate framework for the issue ^[17].

The 2023 Meeting of the High Contracting Parties, held from 15 to 17 November 2023, gave the GGE its current mandate: to formulate by consensus "a set of elements of an instrument, without prejudging its nature," to submit a report to the Seventh Review Conference, to complete its work "as soon as possible, preferably before the end of 2025," and to meet for ten days a year ^[18]. The Group met on 3 to 7 March and 1 to 5 September 2025, and its 2026 sessions were scheduled for 2 to 6 March and 31 August to 4 September ^[19][20].

The chair's rolling text dated 5 June 2026 records provisional consensus on 22 formulations, on the understanding that nothing is agreed until everything is agreed ^[5]. Its opening formulation characterizes a LAWS as a combination of one or more weapons and functionally integrated technological components that can identify, select and engage a target without intervention by a human operating the system, and clarifies that pre-programming a set of potential targets does not remove a system from that description ^[5]. Later paragraphs prohibit LAWS whose effects in attack cannot be anticipated and limited; require "human judgement and control" while stating that this "does not imply constant direct human control"; require that systems be deactivatable or neutralizable in a timely manner and that target and mission parameters not be substantially modified while a system is active unless a human executes or authorizes the change; and ask states to prevent and mitigate unintended bias, including bias in data sets and artificial intelligence models, and separately to prevent and mitigate automation bias ^[5]. According to the Campaign to Stop Killer Robots, 42 states (39 High Contracting Parties and three observers) declared in a joint statement delivered by Brazil in September 2025 that the rolling text was a sufficient basis for negotiations and that they were ready to move ahead toward them ^[21].

General Assembly resolutions

Frustration with the CCW's consensus rule pushed the issue into the General Assembly, where majority voting applies.

Scroll sideways for more →

Belarus, India, Mali and the Russian Federation voted against the 2023 resolution; Belarus, the Democratic People's Republic of Korea and the Russian Federation voted against the 2024 one, with China, India, Israel and Türkiye among the abstainers ^[22][23]. The First Committee approved the 2025 text, draft resolution A/C.1/80/L.41, as a whole at its 26th meeting on 6 November 2025 by 156 votes to 5 with 8 abstentions ^[24].

The Secretary-General's report of 1 July 2024, prepared under resolution 78/241, consolidated submissions from states, international organizations, the ICRC, civil society, academia and industry ^[3]. In its conclusions António Guterres wrote that "machines that have the power and discretion to take human lives are politically unacceptable and morally repugnant, and should be banned by international law," and reiterated his call for the conclusion by 2026 of a legally binding instrument to prohibit lethal autonomous weapons systems that function without human control or oversight and that cannot be used in compliance with international humanitarian law, and to regulate all other types of autonomous weapons systems, adding that "the autonomous targeting of humans by machines is a moral line that must not be crossed" ^[3]. This two-tier structure of prohibitions plus regulations has become the standard framing of what a treaty might contain.

Human rights bodies reached the issue earlier. In April 2013 the UN Special Rapporteur on extrajudicial, summary or arbitrary executions, Christof Heyns, recommended that the Human Rights Council call on all states to declare and implement national moratoria on at least the testing, production, assembly, transfer, acquisition, deployment and use of lethal autonomous robotics until an internationally agreed framework on their future had been established ^[25].

United States policy

DoD Directive 3000.09 was first issued in November 2012 and reissued on 25 January 2023 over the signature of Deputy Secretary of Defense Kathleen Hicks ^[1][2]. Contrary to frequent reporting, it does not prohibit developing or fielding autonomous weapons ^[2]. It requires that autonomous and semi-autonomous weapons be designed to allow commanders and operators to exercise "appropriate levels of human judgment over the use of force," that systems undergo verification, validation, and realistic developmental and operational testing, that engagements terminate or seek further operator input if they cannot be completed within intended constraints, and that human-machine interfaces be readily understandable to trained operators ^[1].

Its most distinctive feature is a senior review. Autonomous weapon systems outside the exempted categories must be approved by the Under Secretary of Defense for Policy, the Under Secretary for Research and Engineering, and the Vice Chairman of the Joint Chiefs of Staff before formal development, and by the Under Secretary for Policy, the Under Secretary for Acquisition and Sustainment, and the Vice Chairman before fielding ^[1]. Any of those officials may request a Deputy Secretary of Defense waiver in cases of urgent military need ^[1]. The 2023 reissue also established an Autonomous Weapon Systems Working Group to support the reviews and tied the use of AI in these systems to the DoD AI Ethical Principles adopted in February 2020 ^[1].

Congress has attached reporting requirements. Section 251 of the FY2024 National Defense Authorization Act requires notification of any change to the directive within 30 days, and Section 1066 of the FY2025 Act requires an annual report on the approval and deployment of lethal autonomous weapon systems through the end of 2029 ^[2]. Separately, the State Department issued a Political Declaration on Responsible Military Use of Artificial Intelligence and Autonomy on 9 November 2023, which 58 states had endorsed as of 27 November 2024 ^[2]. CRS notes that roughly 30 countries and 165 non-governmental organizations have called for a preemptive ban, and that the United States does not support one ^[2].

Project Maven and the targeting-support distinction

Project Maven is frequently invoked in autonomous weapons debates although it is not a weapon. It was established by a Deputy Secretary of Defense memorandum of 26 April 2017 as the Algorithmic Warfare Cross-Functional Team, to accelerate DoD's integration of big data and machine learning and to turn the department's volume of data into actionable intelligence and insights at speed ^[26]. Its first task was to augment or automate processing, exploitation and dissemination of full-motion video from tactical and mid-altitude drone sensors in support of the Defeat-ISIS campaign, initially by providing computer vision algorithms for object detection, classification and alerts, integrated with programs of record in 90-day sprints ^[26]. The system flags and sorts imagery to reduce the analytical burden on human analysts, and the memorandum confers no authority to engage targets ^[26].

Commercial AI suppliers began setting their own limits during the same period. On 7 June 2018 Google chief executive Sundar Pichai published a set of AI principles stating that the company would not design or deploy AI for "weapons or other technologies whose principal purpose or implementation is to cause or directly facilitate injury to people," nor for surveillance violating internationally accepted norms ^[27]. That post now carries an update notice pointing to revised principles published on 4 February 2025 by James Manyika and Google DeepMind chief executive Demis Hassabis, which are organized around bold innovation, responsible development and deployment, and collaborative progress, and which do not restate the earlier list of applications the company would not pursue ^[28].

Arguments for and against

The case for permitting autonomous weapons is usually made on precision and human fallibility grounds. Ronald Arkin of the Georgia Institute of Technology argued in a 2007 technical report that, while he did not believe any unmanned system could be perfectly ethical in the battlefield, autonomous robots could eventually perform more ethically than human soldiers are capable of. He listed six reasons: they need not have self-preservation as a foremost drive and can therefore act conservatively under target uncertainty; sensors may eventually exceed human battlefield perception; they can be built without emotions such as anger or fear that distort judgment; they avoid "scenario fulfillment," the premature cognitive closure implicated in the 1988 downing of an Iranian airliner by the USS Vincennes; they can integrate more information from more sources faster than a person can in real time; and, operating alongside human troops, they could monitor and report ethical infractions ^[29]. The US government has made a related institutional argument, that automated identification, tracking, selection and engagement can strike military objectives more accurately and with less collateral damage ^[2].

The case against runs along four lines. The first is legal: international humanitarian law requires combatants carrying out a specific attack to make context-dependent, evaluative judgments about distinction, proportionality and precautions, and the ICRC argues that the way autonomous weapons function, with the user not choosing the specific target or the precise time or location of a strike, makes that difficult ^[30]. The second is the responsibility gap. Robert Sparrow argued in a 2007 Journal of Applied Philosophy paper that neither the designers, nor the commanding officer, nor the machine can justly be held responsible for an atrocity committed by an autonomous weapon, and that because just war theory requires someone to be justly responsible for deaths in war, deploying such systems would be unethical ^[31]. Human Rights Watch and Harvard Law School's International Human Rights Clinic developed the same point in a 2015 report, concluding that criminal liability fails because machines lack the requisite mental state and commanders rarely meet the knowledge threshold, while civil liability fails because of military and contractor immunities and the difficulty of proving a design defect in autonomous software ^[32]. The third is dehumanization: the ICRC describes autonomous weapons as reducing, or risking removing, human agency in decisions to kill, injure and destroy, which it calls a dehumanizing process that undermines shared humanity, and the Campaign to Stop Killer Robots argues that machines cannot recognize people as people and that allowing systems to target people would reinforce or exacerbate existing structures of inequality, because prejudices live in data sets, categories, labels and algorithms ^[30][33]. The fourth is escalation and proliferation risk, which the General Assembly itself has flagged, citing an emerging arms race, lowered thresholds for conflict, miscalculation, and diffusion to unauthorized recipients and non-state actors ^[7].

Battlefield reality and recent developments

The war in Ukraine is often described as the first conflict where autonomous weapons were used at scale, and the evidence for that framing is thinner than the rhetoric. A March 2025 CSIS analysis by Kateryna Bondar found that autonomy, in the US military's sense of a system accomplishing goals independently or with minimal supervision in complex and unpredictable environments, was not yet present on the battlefield in Ukraine; that although such systems may operate without direct human control they typically do not perform the entire process of finding, selecting and engaging targets independently; and that human oversight remained critical in engagement decisions ^[34]. What had spread was partial autonomy in specific functions, above all autonomous navigation, including the last-mile or terminal navigation that keeps a strike on course when the radio link is jammed. Bondar reported that drones equipped with autonomous navigation raise the target engagement success rate from around 10 to 20 percent to around 70 to 80 percent, roughly a three to fourfold improvement ^[34]. She described the ZIR autonomy kit, a module about the size of a bar of soap combining a camera and computer with a pretrained model that automatically detects targets, and, in a separate case study, the Skydio X10D, a jamming-resistant reconnaissance drone used primarily for tactical surveillance whose automatic target recognition draws on deep learning and computer vision models refined for both electro-optical and infrared imagery ^[34]. Autonomy of this kind sits inside DoD's semi-autonomous category, which expressly covers terminal guidance that homes in on targets an operator has selected, but it shortens the interval in which a human could change their mind ^[1].

The policy questions have also begun to reach commercial AI developers directly. On 27 February 2026 President Donald Trump directed federal agencies to cease all use of Anthropic's technology, and the Secretary of Defense subsequently directed the Department of Defense to designate the company a supply-chain risk to national security, to bar defense contractors and suppliers from working with it, and to set out a transition period of up to six months ^[35]. According to CRS the dispute followed Anthropic's refusal, after the Pentagon asked AI companies to allow use of their models for all lawful purposes, to permit two categories of use: mass domestic surveillance and fully autonomous weapon systems. Chief executive Dario Amodei said that autonomous weapon systems "may prove critical for our national defense. But today, frontier AI systems are simply not reliable enough to power fully autonomous weapons" ^[35]. Anthropic filed a civil complaint in the Northern District of California and a petition for review in the D.C. Circuit on 9 March 2026; the district court issued a preliminary injunction in the company's favor on 26 March, and the court of appeals denied Anthropic's motion for a stay on 8 April, which CRS describes as undoing the lower court's injunction ^[35]. CRS notes that DoD is not publicly known to use any frontier AI model inside an autonomous weapon system, and that some members of Congress have called for legislation on departmental use of AI and autonomous weapons ^[35]. The dispute made concrete a question that the Geneva process has debated abstractly for a decade: who decides how much autonomy is acceptable, the state that fields the weapon or the company that supplies the model.

See also

• Human in the loop
• AI ethics
• Drone AI
• Automation bias
• AI governance
• Explainable AI

References

1. ^US Department of Defense, DoD Directive 3000.09, "Autonomy in Weapon Systems," effective 25 January 2023 (reissues the directive of 21 November 2012). web.archive.org/...300009p.pdf
2. ^Congressional Research Service, "Defense Primer: U.S. Policy on Lethal Autonomous Weapon Systems," IF11150, updated 26 March 2026. everycrsreport.com/...IF11150
3. ^United Nations Secretary-General, "Lethal autonomous weapons systems: Report of the Secretary-General," A/79/88, 1 July 2024. docs.un.org/...88
4. ^International Committee of the Red Cross, "ICRC position on autonomous weapon systems," 12 May 2021. icrc.org/...icrc-position-autonomous-weapon-systems
5. ^Group of Governmental Experts on LAWS, "Rolling text, status date: 05 June 2026," UN Office for Disarmament Affairs documents library. docs-library.unoda.org/...g_Text_-_5_June_2026.pdf
6. ^UN General Assembly resolution 78/241, "Lethal autonomous weapons systems," adopted 22 December 2023. docs.un.org/...241
7. ^UN General Assembly resolution 79/62, "Lethal autonomous weapons systems," adopted 2 December 2024. docs.un.org/...62
8. ^UN General Assembly resolution 80/57, "Lethal autonomous weapons systems," adopted 1 December 2025. docs.un.org/...57
9. ^Human Rights Watch and Harvard Law School International Human Rights Clinic, "Losing Humanity: The Case against Killer Robots," 19 November 2012. hrw.org/...case-against-killer-robots
10. ^Article 36, "Killer Robots: UK Government Policy on Fully Autonomous Weapons," April 2013. article36.org/...Policy_Paper1.pdf
11. ^RTX (Raytheon), "Phalanx Close-In Weapon System" product page. rtx.com/...phalanx-close-in-weapon-system
12. ^Israel Aerospace Industries, "HARPY" product page. iai.co.il/...harpy
13. ^Israel Aerospace Industries, "HAROP" product page. iai.co.il/...harop
14. ^UN Security Council, "Letter dated 8 March 2021 from the Panel of Experts on Libya established pursuant to resolution 1973 (2011) addressed to the President of the Security Council," S/2021/229, 8 March 2021, paragraphs 63-64 and table 4. docs.un.org/...229
15. ^Zachary Kallenborn, "Was a flying killer robot used in Libya? Quite possibly," Bulletin of the Atomic Scientists, 20 May 2021. thebulletin.org/...ot-used-in-libya-quite-possibly
16. ^Fifth Review Conference of the High Contracting Parties to the CCW, final document, CCW/CONF.V/10, 23 December 2016 (Decision 1 establishing the GGE). docs.un.org/...10
17. ^Meeting of the High Contracting Parties to the CCW, final report, CCW/MSP/2019/9, 13 December 2019, Annex III (guiding principles). docs.un.org/...2019
18. ^Meeting of the High Contracting Parties to the CCW, final report, CCW/MSP/2023/7, 23 November 2023 (current GGE mandate). docs.un.org/...2023
19. ^GGE on LAWS, provisional programme of work, CCW/GGE.1/2025/2, 17 December 2024. docs.un.org/...2025
20. ^GGE on LAWS, provisional agenda, CCW/GGE.1/2026/1, 12 December 2025. docs.un.org/...2026
21. ^Campaign to Stop Killer Robots, "Cross-regional group of states ready to start treaty negotiations on autonomous weapons," 8 September 2025. stopkillerrobots.org/...r-2025-gge-joint-statement
22. ^UN General Assembly, verbatim record, A/78/PV.50 (Resumption 1), December 2023 (vote on draft resolution XXXIII, adopted as resolution 78/241). docs.un.org/...PV.50(Resumption1)
23. ^UN General Assembly, verbatim record, A/79/PV.44, 2 December 2024 (vote on draft resolution XLIII, adopted as resolution 79/62). docs.un.org/...PV.44
24. ^UN General Assembly First Committee, "List of proposals contained in the report of the First Committee: Note by the Secretariat, Addendum," A/C.1/80/INF/3/Add.1, 18 November 2025 (voting record for A/C.1/80/L.41, adopted as a whole 156-5-8 at the 26th meeting, 6 November). docs.un.org/...Add.1
25. ^Christof Heyns, "Report of the Special Rapporteur on extrajudicial, summary or arbitrary executions," A/HRC/23/47, 9 April 2013. docs.un.org/...47
26. ^Deputy Secretary of Defense Robert O. Work, memorandum, "Establishment of an Algorithmic Warfare Cross-Functional Team (Project Maven)," 26 April 2017. govexec.com/...ment_of_the_awcft_project_maven.pdf
27. ^Sundar Pichai, "AI at Google: our principles," Google blog, 7 June 2018. blog.google/...ai-principles
28. ^James Manyika and Demis Hassabis, "Responsible AI: Our 2024 report and ongoing work," Google blog, 4 February 2025. blog.google/...onsible-ai-2024-report-ongoing-work
29. ^Ronald C. Arkin, "Governing Lethal Behavior: Embedding Ethics in a Hybrid Deliberative/Reactive Robot Architecture," Georgia Institute of Technology Technical Report GIT-GVU-07-11, 2007. sites.cc.gatech.edu/...formalizationv35.pdf
30. ^International Committee of the Red Cross, "What you need to know about autonomous weapons," 26 July 2022. icrc.org/...you-need-know-about-autonomous-weapons
31. ^Robert Sparrow, "Killer Robots," Journal of Applied Philosophy 24(1), 2007, pp. 62-77. doi.org/...j.1468-5930.2007.00346.x
32. ^Human Rights Watch and Harvard Law School International Human Rights Clinic, "Mind the Gap: The Lack of Accountability for Killer Robots," 9 April 2015. hrw.org/...lack-accountability-killer-robots
33. ^Campaign to Stop Killer Robots, "Facts about autonomous weapons." stopkillerrobots.org/...s-about-autonomous-weapons
34. ^Kateryna Bondar, "Ukraine's Future Vision and Current Capabilities for Waging AI-Enabled Autonomous Warfare," Center for Strategic and International Studies, 6 March 2025. csis.org/...s-waging-ai-enabled-autonomous-warfare
35. ^Congressional Research Service, "Pentagon-Anthropic Dispute over Autonomous Weapon Systems: Potential Issues for Congress," IN12669, updated 21 April 2026. everycrsreport.com/...IN12669
36. ^Campaign to Stop Killer Robots, "About us." stopkillerrobots.org/about-us