US-Russia Controversy over Joint Obstruction of AI Weapons System Regulation: Background and Implications

Executive Summary

During the UN General Assembly, it was confirmed that the United States and Russia jointly blocked the introduction of a clause mandating human intervention in AI weapons system targeting decisions. This is a combined result of the Trump administration's skepticism toward international AI oversight and Russia's distrust of Western-led multilateral norms, reflecting the structural position shared by both countries as permanent members of the UN Security Council and nuclear-armed states in disarmament norms. Meanwhile, 22 countries, including European nations, have separately adopted a declaration on the principle of human control in an attempt to fill the normative gap. Consequently, AI governance is fragmenting into three tracks: the US-China bilateral channel, the US-Russia obstruction track, and the multilateral declaration of middle powers. Since the formation of official multilateral norms is highly likely to stagnate, South Korea needs to prioritize participating in SIPRI-style informal working-level consultative bodies and establishing its own AI verification procedures. In a situation where the United States adopts a dual approach—blocking regulations with Russia while maintaining an accident notification channel with China—South Korea must assess both its scope for independent response and the limits of its bargaining power.

I. Issue Analysis

US-Russia Controversy over Joint Obstruction of AI Weapons System Regulation: Issue Analysis

1. Background and Progress of the Issue

This issue first came to light through on-site reporting at the UN General Assembly by the Italian newspaper La Repubblica, which cited a Washington Post report [1]. The core of the report is that the United States and Russia jointly blocked the introduction of international restrictions on the use of AI in weapons systems [1]. In particular, it highlighted that a proposal to mandate that final decisions on targets suggested by AI be made by humans rather than computers was defeated [1].

This controversy is not an isolated incident. It occurred in the same context as the US skepticism toward AI regulation that ran through the entire week of the UN General Assembly in September. President Trump defined demands for international oversight of AI as a "globalist ploy" in his speech at the UN General Assembly [4]. He also declared that the US government would use the term "superintelligence" instead of "artificial intelligence" in official documents [4]. Around the same time, Russian Foreign Minister Sergey Lavrov spoke at the UN, accusing Europe of obstructing peace regarding the war in Ukraine [1]. The fact that the two countries shared interests on a separate track—AI weapons regulation—adds weight to this report.

During the same UN General Assembly, 22 countries, including European nations, adopted a declaration stating that AI must remain "under human direction, supervision, and control" [10]. This declaration proposed the creation of a global oversight framework and warned that the rapid development of advanced AI models poses serious risks to safety and security if not properly managed [10]. The obstruction by the US and Russia occurred at the exact point of direct conflict with these multilateral attempts at normalization.

2. Current Situation

Around the time of the UN General Assembly, discussions on AI governance have split into two opposing trends. One is the US-China bilateral track. Following the Trump-Xi summit in Washington, the two countries agreed to establish a channel for mutual notification in the event of AI accidents, with the next round of consultations scheduled for November [16]. UN Secretary-General António Guterres has publicly urged the two major AI powers, the US and China, to initiate dialogue [18]. However, EAI analysis assesses that this US-China agreement is "closer to a gesture for domestic politics on both sides rather than a normative achievement with substantive binding force" [9]. Furthermore, it points out that this hotline is "merely a national security crisis management channel and does not imply an expansion of commercial AI cooperation" [9].

The other is the weapons system regulation track, to which the joint US-Russia obstruction belongs. On September 9–10, the Stockholm International Peace Research Institute (SIPRI) held closed-door discussions on state-industry cooperation for the lawful development and use of military AI, inviting officials from ministries of defense and foreign affairs, the International Committee of the Red Cross (ICRC), NATO, and the UN Institute for Disarmament Research (UNIDIR) [8]. While this shows that efforts to establish norms for military AI are continuing at the working level, it stands in stark contrast to the fact that the core clause retaining final targeting authority for humans was defeated due to US-Russia collusion [1][8].

A case separately analyzed by EAI underscores the urgency of this issue. This spring, the US military prepared an interdiction and seizure operation against a Chinese vessel in Middle Eastern waters based on an intelligence report co-authored by an AI chatbot, only to abort the operation at the last minute with fighter jets already scrambled [6]. EAI points out that "the source of the error was not deception by the adversary, but the US military's own intelligence production process," and assesses that "this demonstrates that the key to crisis stability lies not in notification channels, but in the timing of human verification procedures" [6]. In other words, while the principle of human control is being discussed formally, a verification gap has already manifested in actual military operations.

3. Key Actors and Positions

United Statesis a key party to this issue. The Trump administration has prioritized maintaining the pace of AI development as a national strategy and views international oversight frameworks as mechanisms that restrict US technological leadership [4]. In his UN speech, President Trump stated, "Instead of suppressing AI, the United States will continue to encourage its development" [17]. This position forms a policy stance consistent with the reports of blocking international restrictions on military AI. At the same time, the United States is exhibiting a dual attitude: establishing an accident notification channel with China [16] while sharing interests with Russia in blocking weapons regulation.

Russiahas been identified alongside the United States as a co-actor in blocking the regulation [1]. Foreign Minister Lavrov's remarks at the UN show a contradictory approach: while confronting the West over the war in Ukraine, Russia is engaging in pragmatic cooperation with the United States on the separate issue of AI militarization [1]. The Peterson Institute for International Economics (PIIE) notes the potential for China and Russia to cooperate in advanced AI domains, while predicting that the United States, the West, and Ukraine will also cooperate on AI-based weapons systems [13]. This suggests that the alignments surrounding AI militarization do not form a simple US-Russia versus the West dichotomy, but rather a complex landscape of interests reshaped on an issue-by-issue basis.

European Countriesare the camp leading the 22-country declaration and demanding the normalization of the human control principle [10]. Given that their demands were effectively thwarted by the joint US-Russia obstruction, Europe can be seen as a direct stakeholder and affected group in this matter.

Chinadoes not appear directly in this report, but is exercising separate bargaining power on the parallel US-China track. While China seeks to use AI safety discussions as a means to secure global rule-making authority, the United States addresses safety issues solely within the framework of maintaining its competitive advantage over China, leading to assessments that the two sides pursue fundamentally different goals [9].

4. Key Issues

The first issue is the hollowing out of the principle of human control. The clause mandating that humans make the final decision on targets suggested by AI has been regarded as the bare minimum point of consensus in military AI governance discussions. The defeat of this clause demonstrates that the principle of "human direction, supervision, and control" in the European-led 22-country declaration can be neutralized in the domain of weapons systems by great power consensus [1][10].

The second issue is the neutralization of multilateral norms through great power collusion. The fact that the United States and Russia—two ideologically and strategically opposed nuclear military powers—formed a common interest on a specific issue to block the introduction of international restrictions reveals the structural limitation that military AI regulation is difficult to enforce solely through the will of middle powers or multilateral organizations.

The third issue is the fragmentation of regulation tracks. While dialogue is underway between the United States and China, albeit in the limited form of an accident notification channel [16], the US and Russia have actively blocked progress on the more fundamental issue of weapons system regulation. This suggests that rather than converging into a single international regime, AI governance is unfolding across multiple tracks fragmented by great power and issue area. For middle powers, this makes it increasingly unclear on which track they can make their voices heard.

II. In-Depth Analysis

US-Russia Controversy over Joint Obstruction of AI Weapons System Regulation: In-Depth Analysis

1. Root Cause Analysis

While the ostensible justification for US-Russia collusion is the debate over technical efficacy, the actual motivation lies in maintaining military superiority. Discussions on establishing norms for Lethal Autonomous Weapons Systems (LAWS) have been ongoing within the framework of the UN Convention on Certain Conventional Weapons (CCW) since the mid-2010s, but have repeatedly failed to reach a binding agreement due to opposition from major military powers. This issue is an extension of that pattern. The United States believes that mandating human intervention in AI targeting decisions would disadvantage it in terms of operational speed compared to competitors. Russia has also already deployed AI-assisted target identification systems on the battlefield in Ukraine and has no incentive to agree to norms that would retroactively restrict them.

The Trump administration's position shows a consistent logical structure on this issue. In his speech at the UN General Assembly, President Trump defined demands for international AI oversight as a "globalist ploy" [4]. He stated that rather than suppressing AI, the United States would continue to develop it [17]. This demonstrates that the rejection of commercial AI regulation and the rejection of military AI regulation operate within the same political frame. In other words, the White House is applying a discourse that defines AI not as an object of regulation but as a source of national competitiveness, without distinguishing between the civilian and military domains [4][17].

Russia's calculations differ, but the outcomes converge. Foreign Minister Lavrov's remarks during the same UN General Assembly accusing Europe of obstructing peace [1] suggest that Russia harbors a fundamental distrust of Western-led multilateral norms. For Russia, AI weapons regulation is highly likely to be perceived as yet another Western-imposed norm that restricts its military freedom of action in conducting the war in Ukraine. The fact that two countries with different underlying interests produced the same outcome—the defeat of regulation—is not a coincidental convergence, but rather a reflection of the structural position shared by both countries as permanent members of the UN Security Council and nuclear-armed states within multilateral disarmament norms.

2. Structural Context

In terms of security structure, this issue exhibits a pattern similar to the collapse of the nuclear disarmament regime. During the Cold War, the United States and the Soviet Union, as two superpowers sharing the overwhelming destructive power of nuclear weapons, adopted a dual approach: cooperating to prevent proliferation while remaining reluctant to reduce their own armaments. The same structure is being replicated in AI weapons systems. While the United States and Russia cooperate to block the formation of international norms regulating the militarization of AI, this is merely passive cooperation aimed at preserving their own freedom of development and deployment. This explains why the United States and Russia are identified as key actors in inter-state norm competition regarding the military use of AI in emerging and non-traditional security domains.

In terms of political structure, the dual-tracking of AI policy within the United States is prominent. EAI analysis has previously pointed out that in US-China relations, "the White House-led cooperation track and the Congress- and regulatory agency-led pressure track are already operating separately" [9]. The joint US-Russia obstruction demonstrates that this dual-track structure manifests similarly in relations with Russia. While the White House shares practical interests with Russia regarding military AI regulation, it maintains a pressure stance, such as semiconductor export controls, in its competition with China [9]. This indicates that US AI diplomacy is being fluidly reorganized based on issue-specific interests rather than strict distinctions between allies and competitors.

In terms of economic structure, the alignment of interests between the defense industry and the AI industry forms the material basis for blocking regulation. The fact that SIPRI held closed-door discussions in September with defense and foreign ministry officials, industry representatives, the ICRC, and NATO on "state-industry cooperation for the lawful development and use of AI in the military domain" [8] suggests a structure where norm formation is effectively decided within state-industry complexes rather than through intergovernmental negotiations. Regardless of the defeat of the clause mandating human intervention in targeting decisions on the UN multilateral stage, working-level cooperative bodies are conducting discussions based on the premise of "lawful use" rather than regulation [8]. This indicates that the pathway for forming international regulatory regimes is shifting from UN General Assembly-style multilateralism to minilateral state-industry network-type governance.

3. Historical Precedents and Comparative Cases

The closest precedent is the repeated failure of the Group of Governmental Experts (GGE) discussions on LAWS under the CCW. Ongoing for over a decade since 2014, these discussions have repeatedly failed to draft a binding treaty due to opposition from leading military technology nations such as the United States, Russia, and Israel. The joint US-Russia obstruction is the latest instance of this failure pattern, representing more advanced information than before in that the actors opposing the regulation have been publicly identified.

The second comparative case is the great power collusion structure seen in the Treaty on the Non-Proliferation of Nuclear Weapons (NPT) regime. During the Cold War, the United States and the Soviet Union cooperated to regulate mutual nuclear testing, but this was largely aimed at solidifying the existing order by preventing the entry of emerging nuclear-armed states. A similar implication can be drawn from the obstruction of AI weapons regulation. Those who benefit from the defeat of the human intervention mandate are the United States and Russia, which already possess advanced AI targeting systems, while relatively latecomer nations will bear the burden of establishing their own standards in a normative vacuum when developing similar systems in the future.

The third case is the recent incident of a false order triggered by AI hallucination in the US military, analyzed by EAI. In this incident, an interdiction and seizure operation against a Chinese vessel in Middle Eastern waters was prepared based on an intelligence report co-authored by an AI chatbot, but was aborted at the last minute after fighter jets had already scrambled [6]. EAI analyzed that "the source of the error was not deception by the adversary, but the US military's own intelligence production process" [6]. This incident can be read as an example where human intervention actually worked to prevent a catastrophic outcome. Nevertheless, the fact that the United States, on the UN stage around the same time, joined Russia in blocking that very mandate for human intervention [1] demonstrates a disconnect between working-level risk perception and norm-level policy choices.

4. Key Variables Shaping Future Developments

The first variable is whether the US-China AI dialogue makes progress. Following the Trump-Xi summit in Washington, the two countries agreed to establish a mutual notification channel for AI accidents, with follow-up consultations scheduled for November [16]. However, this channel is evaluated as "merely a national security crisis management channel and does not imply an expansion of commercial AI cooperation" [9]. Whether this dialogue expands its agenda to include military AI targeting decisions or remains confined to the narrow scope of accident notification will determine the direction of future norm formation.

The second variable is the expansion of the 22-country declaration. Led by European nations, this declaration specified that AI must remain "under human direction, supervision, and control" and proposed the creation of a global oversight framework [10]. The key is whether the number of participating countries expands beyond 22 to form a coalition of a scale capable of pressuring the permanent members of the UN Security Council. Currently, the three major military AI powers—the United States, Russia, and China—are all absent from this declaration, limiting its normative binding force.

The third variable is the self-regulatory achievements of state-industry working-level networks. If the closed-door dialogue among defense ministries, foreign ministries, industry, and international organizations hosted by SIPRI [8] leads to substantive codes of conduct or technical standards, an alternative normative pathway bypassing failed intergovernmental treaties could emerge. However, because this pathway is premised on "lawful use" from the outset [8], it risks producing outcomes closer to justification than regulation.

The fourth variable is the extent to which the actual operation of AI weapons on the battlefield in Ukraine is exposed. As concrete evidence emerges that Russia is already using AI-assisted targeting systems in combat, public backlash against the defeat of the human intervention mandate is likely to grow. This is a potential variable that could lead to pressure for renegotiation on the UN stage.

The implications for South Korea depend on the combination of these four variables. Under the assumption that a binding UN-level military AI norm is unlikely to emerge for the time being, the priority task remains establishing independent verification procedures for AI-generated intelligence entering the ROK-US combined intelligence system. At the same time, multilateral diplomatic efforts must be pursued in parallel, such as participating in middle-power coalition tracks like the 22-country declaration, to fill the normative gap created by US-Russia bilateral collusion.