Russia and the United States blocked an agreement on AI in the military sphere at negotiations in Geneva

Russia and the United States are at the center of a new dispute over the rules for the use of artificial intelligence (AI) in the military sphere. The Washington Post reported that at talks in Geneva, the two countries effectively blocked the preparation of a legally binding agreement on lethal autonomous weapons systems and achieved the removal from the document of several provisions on human control over AI decisions.

Negotiations in Geneva on August 31 - September 4 took place against the backdrop of intensified discussions about the possible consequences of the uncontrolled development of AI, including in the military sphere. The position of Russia and the United States has become a reason for reproaches in the Western media for their alleged reluctance to control the use of autonomous weapons systems and to consolidate human control over decisions made by AI. Reproaches against the United States have become especially acute due to the position of the American authorities on the widespread introduction of AI in the military sphere and the need for further development of the industry, despite possible risks.

However, as Vadim Kozyulin, chief researcher at the Diplomatic Academy of MGIMO of the Russian Foreign Ministry, noted in a conversation with Lenta.ru, in fact, the efforts of Russia and the United States were aimed at trying to correct the document, ridding it of value judgments and bringing it into line with existing norms of international humanitarian law. In his opinion, in reality, a number of experts participating in the discussion tried to introduce into the draft agreement norms on observance of human rights and ethical principles that would dilute its content.

What is known about the negotiations?

The Geneva talks were a key point at the meeting of the GGE, the official negotiating format of the states parties to the Convention on Inhumane Weapons (CCW), created to discuss the regulation and humanitarian aspects of the use of autonomous military equipment.

Negotiations in Geneva did not mark the final stage of discussions on an agreement on AI in the military sphere: during the Seventh Review Conference of the CCW on November 16-20, the participants of the convention will again discuss the possibility of transferring consultations on the settlement into a formal mandate or a legally binding agreement. At the same time, the mandate of the GGE itself expires in 2026, and, as noted by the Russian Ministry of Foreign Affairs (MFA), the September meeting in Switzerland was the last within the three-year work cycle of the current group.

Why did the idea of ​​an agreement on AI in the military sphere arise?

On September 12, the head of one of the largest AI companies, Anthropic, Dario Amodei, became concerned about the accelerating growth of artificial intelligence capabilities and admitted the possibility of swarms of neural network agents taking over the Internet. His idea of ​​a three-stage plan to curb the race of AI systems was supported by the founder of SpaceX and Tesla, Elon Musk.

At the same time, US President Donald Trump responded by announcing a “crazy conspiracy” against AI and called the presence of a “smart president” a condition for effective control, hinting at himself. According to him, the White House, thanks to its settlement powers, managed to keep “all” tech giants, including Dario Amodei, from committing “bad” acts.

Discussions about the role of AI in the military sphere have intensified against the backdrop of the conflict in the Middle East: for example, according to WP, the intelligent data analysis system Palantir's Maven Smart System may have suggested a girls' school in Iran's Minab as a target for a US Army strike in March, resulting in the deaths of more than a hundred people. According to CNN, a report created by a neural network in the spring also almost brought Washington into conflict with China over a proposal to strike a Chinese merchant ship in the Strait of Hormuz.

As Vadim Kozyulin noted, discussions are influenced not only by the pace of development of AI, but also by a significant reduction in the decision-making cycle: in particular, he pointed to the US Army's attempts to introduce operational decision-making from target detection to its destruction in a matter of seconds. The expert also noted the proliferation of swarms of unmanned aerial vehicles with AI, capable of performing a task even if the connection with the operator is lost.

What are Russia's approaches to using AI?

On April 11, Russian President Vladimir Putin, during a meeting on the development of artificial intelligence, called for the development of domestic solutions for national defense. Putin emphasized that in general there are five priority areas in the implementation of AI, including the use of artificial intelligence in the social sphere and economy, as well as in sensitive sectors.

Commenting on the results of the meeting in Geneva, the Russian Foreign Ministry noted the full applicability of current international humanitarian law to the SAS, as well as the adoption by consensus of the participating countries of a working definition of such weapons systems. The department pointed out the need to reduce possible risks associated with their use.

As Vadim Kozyulin notes, during negotiations on the regulation of AI in the military sphere, both Russia and the United States adhere to the point of view that the main subjects regulating the spread and use of technology are, first of all, states. He recalled that existing norms of international humanitarian law provide for the responsibility of specific individuals in the event of the commission of possible war crimes, including those involving the use of artificial intelligence. “The use of AI does not negate the responsibility of the individual, the command and the state for compliance with IHL. The autonomy of individual technical functions does not mean the emergence of “irresponsible” weapons: decisions on the development, deployment and use of such systems are made by people and government agencies,” the political scientist emphasized.

Commenting on the possibility of the emergence of Russian data analysis systems similar to Maven, Kozyulin noted that with a high degree of probability Russia, if not already developed this system, is at least working in this direction. The expert emphasized that this area is one of the main trends in the development of AI around the world.

Will the use of AI in the military sphere be regulated?

According to Vadim Kozyulin, with a high degree of probability, the proposed agreement on regulating AI in the military sphere will contain framework principles based on already existing norms of international humanitarian law.

The political scientist doubted the possibility of creating an organization in the field of regulation of AI, similar to the International Atomic Energy Agency (IAEA). He pointed out the difficulty of creating a mechanism for verifying and controlling the spread of AI technologies due to the potential degree of secrecy of advanced developments, as well as the large number of users of the product.