US and Russia strip human-control safeguards from UN autonomous weapons draft

September 26, 2026 at 06:53 PM2 min

This digest was compiled by AI from multiple sources — links to the originals are below.

US and Russian delegations removed requirements for human review of AI-selected targets, predictability, and reliability from a UN draft on autonomous weapons during closed Geneva talks from August 31 to September 4. The changes were made in a 15-hour session with cameras off and civil society excluded, according to The Washington Post. The agreed text remains non-binding ahead of a November meeting of the Convention on Certain Conventional Weapons.

Key Facts

US and Russian delegations each brought about ten lawyers to the final stage of UN talks on autonomous weapons in Geneva.
Over roughly 15 hours, the two delegations removed requirements for human review of AI-selected targets, predictability, reliability, and ethical considerations from the draft text.
The closed session took place from August 31 to September 4 under the Group of Governmental Experts on lethal autonomous weapons systems.
The agreed text is not legally binding; in November, parties to the Convention on Certain Conventional Weapons will discuss whether to grant an official negotiating mandate.
The Pentagon is reviewing its 2023 directive on autonomous weapons after President Donald Trump ordered an update within 90 days, but no new version had been published by late September.

Closed-Door Negotiations

The final stage of UN talks on lethal autonomous weapons systems took place in Geneva from August 31 to September 4. US and Russian delegations each brought about ten lawyers to the negotiations, according to The Washington Post. During a roughly 15-hour closed session, UN cameras were switched off and civil society representatives were asked to leave the room. One participant described the process as 'death by a thousand paper cuts.' Smaller delegations struggled to keep pace with the speed of the changes, sources told the newspaper.

Removed Safeguards

The draft previously required AI-based systems to operate 'predictably' and 'reliably.' Delegates also removed a provision on considering ethical factors in the use of such weapons. The requirement for human verification of AI-selected military targets before a strike was deleted from the document. The US State Department, the Russian Foreign Ministry, and the UN did not respond to requests for comment.

US Policy Review

The Pentagon is reviewing its 2023 directive on autonomous weapons, which requires human judgment in the use of autonomous and semi-autonomous systems and compliance with international law. President Donald Trump ordered the rules updated within 90 days, but no new version had been published by late September. The US military is expanding AI use in intelligence analysis and target selection. Palantir's Maven system, which processes data from satellites, drones, and radars, became an official Pentagon program in March and can automatically identify potential targets and threats. Anthropic previously demanded Pentagon guarantees that its models would not be used in fully autonomous weapons or mass surveillance.