CNN 5 Things One Thing: The Pentagon vs. AI Power Players: Did Anyone Win?
7 snips
Mar 8, 2026 
Hadas Gold, CNN AI correspondent who covers AI policy and industry, and Dean Ball, former White House science staffer and policy fellow, debate the Pentagon's pact with major AI firms. They unpack why deals with the Defense Department matter, Anthropic’s red lines on weapons and surveillance, OpenAI’s reversal, employee backlash, political influences, and broader policy stakes.
AI Snips
Chapters
Transcript
Episode notes
Pentagon Pushed For Unrestricted Use
- The Pentagon demanded Anthropic allow its models for "all lawful uses" and resisted contractual red lines on autonomous weapons and mass domestic surveillance.
- Anthropic refused those changes citing reliability and legal gaps, triggering a pressure campaign that threatened a supply chain risk designation against them.
OpenAI Reversed Public Support And Took The Deal
- OpenAI publicly supported Anthropic's red lines in the morning but signed a Pentagon contract hours later claiming the same limits were preserved.
- OpenAI employees reacted with frustration, calling the deal rushed and opportunistic, prompting Sam Altman to apologize internally.
Politics Shaped Which AI Vendor Was Favored
- OpenAI's agreement states tools won't be used for domestic surveillance or to direct autonomous weapons, while allowing lawful uses—raising questions why Pentagon accepted OpenAI but not Anthropic.
- Hadas Gold points to politics and personalities, noting OpenAI's leaders' ties to the Trump circle contrasted with Anthropic being labeled "woke."