US, Russia Stripped Safety Clauses From UN AI Weapons Text, Report Says

Diplomats from the United States and Russia moved to delete several key safety provisions from a United Nations negotiating text on lethal autonomous weapons during the closing stage of talks in Geneva earlier this month, according to a newspaper report.

The negotiations ran from August 31 to September 4 and were intended to push forward a framework governing the use of such weapons. On the final day, the two delegations heavily revised the draft, removing language meant to ensure the safe operation of AI-enabled weapons systems.

Among the provisions dropped were requirements that the systems function in a predictable and reliable manner, that ethical considerations guide their deployment, and that humans review military targets identified by AI before any strike is launched.

The report said Washington and Moscow each brought roughly ten lawyers into the revision process — close to double the number of diplomats fielded by other delegations. The pace of the edits left smaller delegations struggling to follow the changes.

The talks are not legally binding, but an agreement among member states could open the way to a binding treaty in future. Countries taking part are due to meet again in Geneva in November to decide whether to press ahead with developing such a treaty.

Experts cited in the report warned that weak rules on lethal AI weapons could allow machines to make life-and-death decisions without human control. The practical consequences, they cautioned, could include more civilian casualties, reduced accountability and a quicker drift toward riskier automated warfare.