US and Russia Remove AI Weapon Safety Rules

The United States and Russia removed key safety provisions from a UN-backed agreement on lethal autonomous weapons systems, including requirements for human review of AI-assigned targets and ethical considerations, during talks in September in Switzerland.

Consensus

• The United States and Russia jointly removed provisions from a UN-backed agreement on autonomous weapons systems.
• The negotiations took place in September in Switzerland, at a UN venue.
• The removed provisions included requirements for human review of AI-assigned military targets.
• The removed provisions also included references to ethical considerations in the use of AI weapons.
• The changes were made over approximately 15 hours during the talks.
• Both US and Russian delegations had around 10 lawyers each, significantly more than other delegations.
• The pace of changes overwhelmed smaller delegations.
• The talks were not legally binding at the time but could lead to a future legally binding treaty.
• A follow-up meeting is planned for November in Geneva to discuss whether the talks should receive an official mandate.

Points of divergence

• The Washington Post reported that the US and Russia removed the requirement for AI systems to function 'predictably' and 'reliably'. — meduza
• The Washington Post reported that US military used a chatbot from Anthropic to identify nearly 1,000 targets within the first day of a war with Iran, and that AI was actively used in military operations in Gaza and Israel. — meduza

Coverage (2 sources)

• The Washington Post: USA and Russia excluded from AI-weapon agreement project a clause about human verification of targets — Meduza
• USA and Russia jointly called for relaxed AI weapon control - Washington Post — DW Главное

Key entities

• USA
• Russia
• UN
• The Washington Post
• Switzerland
• Washington Post
• WP