Pentagon chose OpenAI models for use in AI, replacing Anthropic

Agreement Between OpenAI and the Pentagon: A New Phase in the Conflict with Anthropic

On Friday evening, Sam Altman—CEO of OpenAI—announced a settlement with the U.S. Department of Defense to adapt OpenAI solutions for the agency’s needs. This was a response to a recent dispute that arose because Anthropic refused to allow the Pentagon to use its developments without strict ethical constraints.

What Happened
1. Contract with the Pentagon

- OpenAI and the Department of Defense agreed on terms for using the company’s technologies in classified systems.

- Altman emphasized that throughout all stages of interaction, the DoD showed deep respect for security issues and a willingness to cooperate.

2. Reason for the Conflict with Anthropic

- Anthropic declined an agreement worth $200 million because it did not agree to remove its ethical restrictions.

- As a result, the agency added Anthropic to a list of “suspicious” suppliers, fearing a national security threat. All government entities were ordered to stop using the Claude chatbot and find an alternative within six months.

Key Terms of Cooperation with OpenAI
- Ban on Mass Surveillance: AI will not be used for systems that conduct mass surveillance of U.S. citizens.

- Limitation on Autonomous Weaponry: Fully autonomous combat systems capable of killing people without human involvement are prohibited.

Altman stated that OpenAI would ensure the Department of Defense adheres to these principles. A separate team of OpenAI specialists would oversee compliance.

Call for a Unified Approach
Sam Altman asked the Pentagon to extend similar cooperation terms to all AI companies in order to eliminate contradictions with other market participants.

Anthropic’s Reaction
Anthropic representatives expressed deep disappointment with the Pentagon’s decision to list their company as unreliable. The company reserved the right to appeal this decision in court.