The United States and Russia worked to weaken proposed international safeguards for lethal autonomous weapons during United Nations negotiations in Geneva this month, removing provisions that would have required greater reliability, ethical consideration and human oversight of artificial intelligence systems used in warfare, according to people familiar with the closed-door discussions and documents reviewed by The Washington Post.
The changes were made during roughly 15 hours of negotiations on the final day of the talks. Three people familiar with the proceedings said US and Russian diplomats removed language requiring autonomous weapons systems to operate in a “predictable” and “reliable” manner, as well as a provision requiring ethical considerations when AI weapons were used. They also removed a requirement that humans review military targets generated by AI before a strike.
The negotiations took place behind closed doors, with United Nations cameras switched off and civil society observers excluded. The US and Russian delegations deployed teams of about 10 lawyers each, nearly twice the diplomatic presence of other delegations, according to the three people. The pace of changes left smaller delegations struggling to follow the proceedings.
“It was death by a thousand paper cuts,” one of the people said.
The US State Department, Russian Foreign Ministry and United Nations did not respond to requests for comment.
The talks represent the most advanced international effort to establish rules governing the battlefield use of lethal autonomous weapons. The current negotiations are non-binding, but participants could eventually establish a mandate for a legally binding treaty. Nations are due to meet again in Geneva in November to consider whether the process should move towards an official mandate or potentially a binding agreement.
Verity Coyle, deputy director for arms in the crisis, conflict and arms division at Human Rights Watch, said weak international rules could have significant consequences. “It could mean machines can make life-and-death decisions without human control,” she said. “The practical impact is stark: more civilian harm, less accountability and a faster slide into riskier, automated warfare.”
The negotiations come as AI is increasingly being used in military and security operations. The US military used Anthropic’s chatbot to identify nearly 1,000 targets for strikes during the first 24 hours of the war with Iran, according to The Post, while Israel has made extensive use of AI in its military operations in Gaza, The Post has reported.
AI companies have also disclosed cases involving autonomous hacking of other organisations’ computer systems, including US government websites. Reports that actors in Yemen and Iran have used chatbots in efforts to develop weapons or target the US military have added to concerns over the technology’s military applications.
The diplomatic dispute is mirrored by a debate inside the United States. The Pentagon is rewriting its 2023 directive governing lethal autonomous weapons after President Donald Trump issued a national security memorandum in June requiring an update within 90 days. As of late September, the revised directive had not been released.
The 2023 directive, adopted during the Biden administration, required human judgement to check autonomous and semi-autonomous weapons systems and required operators to comply with the laws of war and applicable treaties.
Nicole van Rooijen, executive director of Stop Killer Robots, said Washington and Moscow opposed legally binding rules and had disproportionate influence over the negotiations. Richard Lennane, disarmament adviser to the International Committee of the Red Cross legal team, said international treaties could establish norms that influence state behaviour.
“It’s not about punishment, it’s about setting norms and influencing behavior,” Lennane said. “We believe that a treaty regulating autonomous weapons could be similarly effective.”
