Anthropic pitched Pentagon's $100M drone swarm challenge

Anthropic's feud with the Pentagon just took an ironic twist.

According to new reports, Anthropic secretly submitted a pitch earlier this year for a $100 million Pentagon prize challenge aimed at developing voice-controlled, autonomous drone swarm technology.

The challenge was launched by US Special Operations Command and the Defense Innovation Unit, designed to build swarms of autonomous drones capable of coordinating across air and sea.

Intent translation, not target selection

For a company built on AI safety, pitching drone swarms sounds like a sharp 180. But people familiar with the pitch say Anthropic didn't see it as crossing its own ethical lines.

The plan was to use Claude as a high-level mission controller. A human commander would speak an objective, and Claude would translate that voice intent into structured digital commands to orchestrate the swarm's flight patterns.

Crucially, Claude wouldn't touch targeting or weapons release. Humans would maintain constant oversight and could kill the mission at any time.

However, the Pentagon's ambitions go much further. Defense officials stated in January that the contest is explicitly meant to enhance offensive capabilities, and that human-machine interfaces are expected to drive the lethality and effectiveness of the weapons. Later phases plan to cover target awareness and operations from launch to termination.

Out of the running, then blacklisted

Anthropic wasn't selected in the first round, and things unraveled quickly.

Last Friday, Defense Secretary Pete Hegseth ordered the Pentagon to bar its contractors and partners from doing any commercial business with Anthropic, following disputes over autonomous military limits.

Meanwhile, competitors jumped right in. A joint bid from SpaceX and xAI made the cut. Defense firms partnering with OpenAI, including autonomous systems firm Applied Intuition, were also selected to handle mission-control voice translation.

Right after Anthropic was sidelined, Sam Altman announced a fresh deal with the Pentagon to deploy OpenAI models on classified military cloud infrastructure.

Why it matters

Anthropic is learning the hard way that walking a moral tightrope in defense tech is brutally difficult. The company tried to split the difference by offering swarm coordination while refusing to touch lethal targeting, but the military ultimately wants tools optimized for operational effectiveness, not safety lectures.

While Anthropic spent months negotiating red lines with defense officials, OpenAI and xAI moved swiftly to swallow the market share. By agreeing to classified cloud deployments and offering the military the exact interfaces it needed, rivals proved far more pragmatic.

We think this boycott could severely hurt Anthropic. Losing a $100 million drone contest is a minor sting, but getting completely blacklisted across all defense contractors cuts the company off from billions in sovereign cash and leaves national security AI entirely in the hands of its fiercest competitors.