Overview
- OpenAI confirmed an agreement to run its models inside the Defense Department’s classified network with stated red lines on mass domestic surveillance, controlling autonomous weapons, and automated high‑stakes decisions, and it says it can terminate the deal on breach.
- CEO Sam Altman said OpenAI will add contract language specifying that intelligence agencies such as the NSA cannot use its services without a separate amendment.
- After Anthropic declined to permit use for “all lawful purposes,” the administration labeled the company a supply‑chain risk and ordered federal users to transition off its technology over six months.
- Despite the designation, multiple reports say U.S. forces continued to use Anthropic’s Claude during recent operations, including an Iran strike, reflecting an ongoing transition period.
- Anthropic signaled a legal challenge to the designation, while user backlash to OpenAI’s deal saw public ChatGPT cancellations and a surge in Claude downloads to the top of U.S. app‑store charts.