GENEVA — The United States and Russia removed several proposed safety and human-oversight provisions from a draft United Nations framework on lethal autonomous weapons during negotiations in Geneva earlier this month, according to The Washington Post and Human Rights Watch.
The talks were held under the UN Convention on Certain Conventional Weapons (CCW) Group of Governmental Experts on Lethal Autonomous Weapons Systems (LAWS). The negotiations, which ran from late August to September 4, were part of years of UN efforts to establish international rules for weapons that can identify, select and engage targets with limited or no direct human intervention.
Key safeguards removed
During about 15 hours of negotiations on the final day, U.S. and Russian diplomats revised the draft and removed a provision that would have required human review of AI-selected military targets before a strike.
According to The Washington Post, the delegations also removed language requiring autonomous weapons to operate in a “predictable” and “reliable” manner. A provision calling for ethical considerations in the use of AI-enabled weapons was also deleted.
Human Rights Watch said other changes weakened or narrowed provisions covering issues including design and development, explainability, traceability and references to international law.
The negotiations took place behind closed doors after UN cameras were turned off and civil-society observers were asked to leave. The U.S. and Russian delegations each had about 10 lawyers, nearly twice the number used by some other delegations. Sources said the revisions moved so quickly that smaller delegations struggled to keep up.
The U.S. State Department, Russian Foreign Ministry and UN did not respond to The Washington Post's requests for comment.
Disagreement over a binding treaty
The negotiations highlighted a long-standing disagreement over how autonomous weapons should be regulated. The United States and Russia have supported non-binding guidelines rather than immediately moving toward a legally binding international treaty.
Human Rights Watch said 76 countries now support beginning negotiations on a legally binding instrument.
The final document nevertheless retains references to human control and judgment and states that autonomous weapons must comply with international humanitarian law. However, rights groups said the final text was substantially weaker than earlier proposals.
AI targeting and civilian casualties
The debate comes as AI-assisted military systems are increasingly being used in modern conflicts.
On February 28, 2026, a U.S. strike hit the Shajareh Tayyebeh girls' elementary school in Minab, Iran, killing approximately 150 to 175 people, including around 120 children, according to Iranian officials and international reporting. The school was located near a military compound.
Investigations raised questions about whether outdated intelligence and the identification of the site as a military target contributed to the strike. The Pentagon opened an investigation. However, available reporting has not established that an autonomous AI system independently selected the school as the target.
AI-assisted targeting has also become a subject of controversy in Gaza. Israel has used systems including Gospel and Lavender to assist with target identification and generation. Reports have raised questions about the extent of human review and civilian casualties, while Israeli officials have said human personnel retain oversight and that the systems are intended to improve targeting accuracy.
Russia and Ukraine have also used AI-enabled technologies in drone warfare, including systems designed to assist drones in identifying or reaching targets. Reports have raised concerns about civilian harm from autonomous or AI-assisted drone operations on both sides.
Political and international response
The Geneva outcome has been criticized by organizations including the Campaign to Stop Killer Robots and Human Rights Watch, which have called for meaningful human control over weapons capable of making life-and-death decisions.
The issue has also gained importance as governments debate wider controls on artificial intelligence. At the UN General Assembly on September 22, 2026, President Donald Trump rejected what he described as a global effort to control AI and said the United States would encourage its development. He also said U.S. government documents would use the term “super intelligence.”
The revised Geneva document is not itself a legally binding treaty. Countries are expected to consider the future of the process at the CCW Review Conference in Geneva from November 16-20, 2026.
The November meeting will help determine whether governments move toward formal negotiations on international rules for lethal autonomous weapons or continue with non-binding measures and existing international humanitarian law.
——— End of Article ———
