Overview
- President Donald Trump ordered federal agencies to halt use of Anthropic’s tools with a six‑month phase‑out as the Pentagon labeled the company a supply‑chain risk.
- Multiple outlets report U.S. Central Command used Anthropic’s Claude for intelligence work and targeting in Iran strikes hours after the directive, highlighting entrenched integrations through partners like Palantir and AWS.
- OpenAI announced an agreement to deploy its models on classified networks under explicit red lines that bar mass domestic surveillance, directing autonomous weapons, and other high‑stakes automated decisions, with controls enforced via cloud deployment and its own safety stack.
- Anthropic refused to remove safeguards for mass domestic surveillance and fully autonomous targeting, called the designation retaliatory, and said it will challenge the action while asserting its stance reflects American values.
- OpenAI’s quick deal drew public and employee backlash as the Pentagon moves to swap out Claude, and analysts warn the transition could disrupt defense workflows and set a precedent for who defines military AI limits.