On 27 February 2026 OpenAI agreed to deploy its models on the Pentagon’s classified network. Sam Altman said the deal keeps prohibitions on domestic mass surveillance and human responsibility for the use of force; a Pentagon official said the terms were the compromise Anthropic had been offered and rejected.

What this evidence bears on

• Supports: OpenAI wrote the same two limits into its own Pentagon deal, so refusing the contract was not the only principled option.

  • A reply in “Anthropic”

• Supports: OpenAI secured limits on mass surveillance and autonomous weapons while still supporting its country’s defence.

  • A reason for “OpenAI”

• Supports: A Pentagon official said these were the terms Anthropic rejected, so the limits may be weaker than they sound.

  • A reply in “OpenAI”