Anthropic CEO Blasts OpenAI-Pentagon Deal as ‘Security Theater’ in Leaked Internal Memo

Anthropic has made Claude available on classified U.S. government networks since June 2024 and secured a Pentagon contract worth up to $200 million in July 2025. The relationship later broke down over two red lines: mass domestic surveillance in the United States and unsupervised autonomous weapons. The dispute centers on whether frontier AI companies can restrict the military’s “lawful uses” of their technology, and has brought the competing safety approaches of OpenAI and Anthropic into public view.

OpenAI reached an agreement with the Pentagon on February 27, 2026, to deploy its technology on classified networks, and strengthened language barring surveillance of Americans on March 2. The value of the agreement was not disclosed. On March 4, The Information reported on a roughly 1,600-word memo that Dario Amodei sent to employees on February 27. In it, he described the technical safeguards as “20% real and 80% security theater” and called Sam Altman’s account “a complete and total lie.”