US and Russia Reportedly Stripped Ethics Rules From Draft Global AI Weapons Treaty
American and Russian diplomats jointly removed key human-control and ethics provisions from a draft UN treaty on autonomous weapons, according to The Washington Post. The changes weakened planned global rules on lethal AI-powered arms.
In early September, hundreds of diplomats gathered in Geneva to draft the first global agreement regulating lethal autonomous weapons systems. According to The Washington Post, citing three people familiar with the talks and internal documents, US and Russian representatives spent 15 hours in a closed session on the final day rewriting the draft, each country bringing roughly 10 lawyers — far more legal staff than other delegations deployed.
What was removed
Sources told the paper that the American and Russian delegations struck language requiring autonomous systems to operate in a "predictable" and "reliable" manner, and deleted a clause obligating parties to consider ethical factors when using AI weapons. They also removed a provision requiring humans to verify AI-identified military targets before a strike was carried out. One person involved described the editing process as "death by a thousand cuts."
Background
The heated Geneva negotiations unfolded as the Pentagon pushes to rapidly integrate AI into its combat strategy, even as debate continues over the existential risks of advanced AI. The Washington Post reported that US forces have already used AI from the company Anthropic to identify and strike nearly 1,000 targets within the first 24 hours of the conflict with Iran. Verity Coyle, deputy arms director at Human Rights Watch, warned that weakening the rules means "machines will be able to make life-and-death decisions without human oversight," which she said would increase civilian harm and accelerate a shift toward riskier, automated warfare.
In January 2023, the US Department of Defense updated its directive on autonomous weapons systems, which had included provisions similar to those now stripped from the UN draft, including a requirement for human review of autonomous and semi-autonomous weapons. In August 2024, UN Secretary-General António Guterres called algorithmic target selection a "moral line that must not be crossed" and urged a binding international treaty by 2026 to fully ban weapons operating without human control. In December 2024, the UN General Assembly adopted its first resolution on lethal autonomous weapons by an overwhelming majority, with only Russia, Belarus and North Korea voting against.
