Particle.news

Pentagon’s Reported Use of Anthropic’s Claude in Maduro Raid Triggers Policy Standoff

The dispute centers on Pentagon demands for fewer safeguards on classified use versus Anthropic’s hard limits on autonomous weapons and mass domestic surveillance.

Overview

  • Multiple outlets report the U.S. military used Anthropic’s Claude during the January operation that captured Nicolás Maduro, with access provided through a Palantir integration.
  • Reporters say the precise functions Claude performed remain undisclosed, though Axios reports it was used during the active raid rather than only in planning.
  • Anthropic says any government use must follow its Usage Policies, which prohibit facilitating violence, weapons development and surveillance, and sources say the company believes those rules were followed.
  • Defense officials are pressing AI vendors to offer models on classified networks with fewer restrictions, and officials are reviewing or threatening a contract with Anthropic worth up to about $200 million, according to WSJ, Reuters, Axios and TechCrunch.
  • Anthropic is reportedly the only major model currently accessible on classified systems via third parties; outlets report no American fatalities in the raid, while Venezuela and Cuba claim dozens of their personnel were killed.