Anthropic CEO slams OpenAI’s military contract as ‘safety theater’

Last week, Anthropic and the U.S. Department of Defense failed to reach an agreement regarding the military’s request for unrestricted access to the company’s artificial intelligence technology. Anthropic, which currently holds a $200 million contract with the military, insisted on protections against the technology being used for domestic mass surveillance or autonomous weaponry. Because the parties could not come to an agreement, the Department of Defense instead struck a deal with competitor OpenAI.

In an internal memo to staff, Anthropic CEO Dario Amodei reportedly criticized OpenAI’s messaging around the new contract as “straight up lies” and “gaslighting.” Amodei characterized OpenAI’s dealings with the Department of Defense as “safety theater”—a term used to describe performative safety measures—rather than a genuine effort to prevent abuses. In the memo, Amodei, who serves as Anthropic’s co-founder and CEO, wrote: “The main reason [OpenAI] accepted [the DoD’s deal] and we did not is that they cared about placating employees, and we actually cared about preventing abuses.”

Public sentiment appears to favor Anthropic following the dispute. ChatGPT uninstalls increased by 295% after OpenAI finalized its agreement with the military. In his memo, Amodei noted that the public and media largely view OpenAI’s deal with the Department of Defense—which he referred to as the “DoW,” or Department of War—as “sketchy or suspicious.” He added that the public instead views Anthropic as the heroes of the situation, pointing to the company’s current rank at #2 in the App Store. While Amodei dismissed the opinions of “Twitter morons,” he expressed concern over ensuring that OpenAI’s messaging does not convince OpenAI’s own employees.

Why it matters

The public spat between Anthropic and OpenAI underscores a growing industry rift over how AI labs should engage with military contracts and the ethical standards governing the use of AI in defense.