The most advanced effort yet to regulate lethal autonomous weapons was weakened from the inside this month, according to a Washington Post report based on three sources and documents the paper reviewed.

During roughly 15 hours of closed-door negotiations in Geneva in early September — with the cameras turned off — U.S. and Russian delegations chipped away at a series of provisions meant to keep humans in control of AI-driven weapons. Deleted language included requirements that such systems function in a "predictable" and "reliable" manner, a clause obliging users to weigh ethical considerations, and a provision requiring humans to verify military targets identified by AI before strikes are carried out.

"It could mean machines can make life-and-death decisions without human control," Verity Coyle, deputy director for arms in the crisis, conflict and arms division at Human Rights Watch, told the Post. "The practical impact is stark: more civilian harm, less accountability and a faster slide into riskier, automated warfare."

The Post reported that Washington and Moscow each deployed roughly ten lawyers — nearly twice the diplomatic presence of other delegations — and moved fast enough to leave smaller delegations struggling to follow. Hundreds of diplomats took part in the talks, which are currently non-binding but could open the door to the broadest global rules yet for AI on the battlefield. States are due back in Geneva in November to decide whether the process can lead to a legally binding treaty.

"They have the power to block any significant progress that is made by the vast majority of states who do want to move forward on a binding treaty," said Nicole van Rooijen, executive director of the Stop Killer Robots coalition. The U.S. State Department, the United Nations and Russia's Foreign Ministry did not respond to the newspaper's requests for comment.

The report lands in the middle of a widening debate over military AI. The Post noted that U.S. forces used an Anthropic chatbot to help identify nearly 1,000 strike targets in the first 24 hours of the war with Iran, while Israel has made extensive use of AI in operations in Gaza. In September, President Donald Trump spoke out against tighter controls on artificial intelligence, telling audiences the world should not "kill the golden goose."

Sources