Dario Amodei publicly apologizes for leaked internal memo criticizing Pentagon and OpenAI deal
Mar 6, 2026
Key Points
- Anthropic CEO Dario Amodei publicly apologizes for a leaked internal Slack post criticizing the Pentagon and a rival AI provider's deal with the Department of Defense.
- The post surfaced on a day Trump called for removing Anthropic from federal contracts and the Defense Secretary designated the company a supply chain risk.
- Senior adviser Gregory C. Allen argues the government should absorb the friction, warning that punishing Amodei over a leaked Slack post would set back Silicon Valley's relationship with national security.
Summary
Dario Amodei issued a public apology in an interview with The Economist for an internal Slack post that criticized the Pentagon and a rival AI company's deal with the Department of Defense, after the post leaked and created a political problem for Anthropic.
Amodei described the Friday in question as among the most disorienting in Anthropic's history. Within a few hours, President Trump posted to have all Anthropic services removed from the federal government, the Secretary of Defense designated Anthropic a supply chain risk, and a competing AI provider struck a deal with the Pentagon that the provider later described as opportunistic. Amodei characterized the post as an unpolished Slack message rather than a considered memo, written in the informal style he uses when communicating with employees. He said he had already apologized to contacts at the Department of Defense and was open to speaking with others in the administration.
In public, Amodei refers only to another AI model provider when discussing the rival deal, without naming the company. The internal Slack post reportedly showed a more direct focus on OpenAI and Sam Altman specifically.
Gregory C. Allen, a senior adviser at the Wadhwani AI Center and the Center for Strategic and International Studies, argued in a separate Stratechery interview that the government should absorb the friction and protect the relationship with Amodei. Allen's case rests on what he sees as a rare achievement: Amodei has brought an entire generation of AI researchers who were skeptical of the national security establishment into active support for defense work, including development of autonomous weapons systems. Allen draws the contrast with Project Maven, where Google's Pentagon collaboration triggered an employee revolt. Anthropic employees, he argues, have moved far enough in that direction that damaging Amodei over a leaked Slack post would set back Silicon Valley's relationship with national security in a way that would be difficult to undo.