At the Pentagon, OpenAI is In and Anthropic Is Out

What It Covers

The Pentagon declared Anthropic a supply chain risk after contract negotiations collapsed over two red lines — mass domestic surveillance and fully autonomous weapons — while OpenAI simultaneously secured a Pentagon deal claiming identical restrictions, raising unresolved questions about whether the agreements are substantively different or politically motivated.

Key Questions Answered

• •The "All Lawful Use" Loophole: The Pentagon demanded AI labs accept an "all lawful use" standard rather than specific prohibitions. Anthropic refused because the U.S. lacks meaningful AI regulation and no national privacy law — meaning activities like purchasing bulk American data from data brokers and feeding it into AI models remain entirely legal, making "lawful" an insufficient safeguard.
• •Supply Chain Risk Designation as a Weapon: The Pentagon's threat to designate Anthropic a supply chain risk — a label historically reserved for Chinese chip suppliers and foreign entities like Kaspersky Lab — represents the most punitive action the U.S. government has taken against a major domestic company in at least a century, with no formal legal proceeding yet initiated beyond a social media post.
• •Data Broker Surveillance Gap: Federal agencies can legally purchase bulk data on millions of Americans from commercial data brokers, then run that data through AI models for analysis. This is functionally equivalent to mass domestic surveillance but falls outside legal definitions of it — making contractual prohibitions on "illegal" surveillance largely meaningless without explicit categorical bans.
• •OpenAI's "Safety Stack" Claim Requires Scrutiny: Sam Altman announced OpenAI would build model-level technical guardrails — a "safety stack" — to prevent misuse inside Pentagon classified networks. Security analysts contacted by the hosts argue these guardrails cannot determine whether data fed into a model was legally obtained, making the technical protection largely performative rather than substantive.
• •Political Alignment as Business Survival Strategy: The Anthropic-Pentagon conflict illustrates a concrete consequence of not cultivating administration relationships. OpenAI's Greg Brockman donated $25 million to Trump's PAC; Tim Cook presented Trump with a trophy. Companies that maintained political alignment secured contracts; Anthropic, perceived as ideologically opposed, faced attempted operational destruction through regulatory designation rather than contractual dispute resolution.