Anthropic co-founder and CEO Dario Amodei has sharply criticized rival OpenAI over its new contract with the U.S. Department of Defense, accusing the company of misleading the public about safeguards on military use of its artificial intelligence systems.

In an internal memo reported by The Information, Amodei described OpenAI’s agreement with the Pentagon as “safety theater” and claimed the company’s public assurances about limits on military applications are “straight up lies.” He argued that OpenAI accepted terms Anthropic had rejected, while trying to present itself as equally cautious.

Anthropic had been negotiating an expanded deal with the Defense Department, building on an existing contract reportedly worth around $200 million. Talks collapsed when the company refused to grant the military unrestricted access to its models. Anthropic sought explicit guarantees that its AI would not be used for domestic mass surveillance or autonomous weapons, and objected to language allowing “any lawful use.”

After those negotiations failed, the Pentagon turned to OpenAI. In a blog post, OpenAI said its defense contract permits use of its systems for “all lawful purposes” and claimed that mass domestic surveillance was excluded because it is not considered lawful. The company said this limitation was made explicit in its agreement.

Amodei, however, told staff that OpenAI’s framing is deceptive, arguing that the “lawful use” standard is far too weak and leaves the door open to future abuses if legal definitions shift. Critics in the AI ethics community have echoed that concern, noting that what counts as “lawful” can change quickly under political pressure or new legislation.

Amodei wrote that Anthropic “actually cared about preventing abuses,” while accusing OpenAI of prioritizing employee optics and public relations. He said OpenAI CEO Sam Altman is “presenting himself as a peacemaker and dealmaker” while accepting terms Anthropic found unacceptable.

The public reaction appears to be tilting in Anthropic’s favor. Third-party analytics cited in coverage of the dispute indicate that ChatGPT app uninstalls spiked after news of OpenAI’s Pentagon deal, suggesting a backlash among some users wary of military AI partnerships.

Amodei told employees that many observers see OpenAI’s arrangement with the Defense Department as “sketchy or suspicious” and cast Anthropic as taking the more principled stance. His memo underscores a growing rift between the two leading AI labs over how far they are willing to go in working with the national security establishment.