Pentagon drops Anthropic, adopts OpenAI under “safety red lines,” signaling a reshaped federal AI stack
Published Feb 28, 2026, 4:17 AM UTC
Key entities
TLDR
Confirmed: The Pentagon ended federal use of Anthropic over security risk concerns and approved OpenAI under defined “safety red lines,” with a reported deal to deploy OpenAI models on a classified DoD network. Near-term: Expect tightened federal vendor criteria, rapid integration tests for OpenAI in defense workflows, and potential legal/market fallout for Anthropic.
Why this matters
Multiple outlets report the U.S. government has ended federal use of Anthropic after Pentagon security risk concerns while moving to adopt OpenAI with explicit safety restrictions, including a reported arrangement to run OpenAI models on a classified DoD network. Anthropic publicly disputes the designation as legally unsound.
What changed
Published 6d after the previous Frontier AI and Model Releases briefing. Source base moved from 6 to 8 sources. Lead sourcing shifted to Auto search: ai.
Topic context
Use this page when you want fast context on OpenAI, Anthropic, Google, xAI, Meta, and major model launches without scanning every product post and policy note yourself. Key angles: openai, anthropic, google, gemini. Latest briefing: Pentagon drops Anthropic, adopts OpenAI under “safety red lines,” signaling a reshaped federal AI stack.
Summary
Multiple outlets report the U.S. government has ended federal use of Anthropic after Pentagon security risk concerns while moving to adopt OpenAI with explicit safety restrictions, including a reported arrangement to run OpenAI models on a classified DoD network. Anthropic publicly disputes the designation as legally unsound.
Sources
Pentagon approves OpenAI safety red lines after dumping Anthropic - Axios
OpenAI hauls in $110B in third round of funding https://www.upi.com/Top_News/US/2026/02/27/opena…
OpenAI strikes deal with Pentagon after Trump orders government to stop using Anthropic https://…
Trump Ends Federal Use of Anthropic’s AI as Pentagon Labels It Security Risk - WSJ
#Trump orders government to stop using Anthropic in battle over AI use The move announced on soc…
OpenAI reaches deal to deploy AI models on U.S. DoW classified network
Anthropic Hits Back After US Military Labels It a 'Supply Chain Risk'
Trump Ends Federal Use of Anthropic’s AI as Pentagon Labels It Security Risk - WSJ