Anthropic had previously bid $100 million on the Pentagon's drone swarm project but was not selected. The victory went to the SpaceX and OpenAI teams.

According to 1M AI News, Anthropic submitted a proposal earlier this year to participate in the Pentagon’s $100 million competition, the Orchestrator Prize Challenge, aimed at developing voice-controlled autonomous drone swarms. The competition, jointly organized by the Defense Autonomous Warfare Group under the Special Operations Command and the Defense Innovation Unit, progresses through five stages, from software development to combat testing. Later stages involve “target perception and sharing” and “from launch to termination.” Anthropic’s plan centers around Claude, converting commander’s intent into digital instructions and coordinating drone formations. It does not involve autonomous targeting or weapon decision-making, with humans maintaining full oversight. The company also proposed establishing joint research projects with the Pentagon to safely develop and evaluate autonomous weapon capabilities.

Anthropic believes this plan does not cross its red line of “not participating in fully autonomous weapons,” as humans can still monitor and terminate the system at any time. However, Anthropic was ultimately not selected, and Bloomberg was unable to confirm the reasons. Selected proposals include a joint bid from SpaceX and xAI, as well as two defense tech companies that list OpenAI as an AI partner (one of which is the autonomous military vehicle contractor Applied Intuition). OpenAI’s technology will be used in the “mission control” phase to help convert voice commands into digital instructions. Hours after the Pentagon announced last Friday that it would ban its contractors from engaging in any commercial activities with Anthropic, OpenAI announced a new agreement with the Department of Defense to use its AI tools on a secure cloud system. Anthropic declined to comment. (Bloomberg)