Anthropic's CEO Dario Amodei has expressed sharp criticism toward OpenAI's Sam Altman. In a staff memo, Amodei characterized OpenAI's recent agreement with the Department of Defense as mere "safety theater." He attributed the differing outcomes between the two companies to their core priorities, writing, "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."

This follows a breakdown in talks between Anthropic and the DoD last week. Despite an existing $200 million contract, Anthropic refused to grant the military unrestricted access to its AI technology without explicit assurances it would not be used for domestic mass surveillance or autonomous weaponry. The Department of Defense subsequently finalized a deal with OpenAI instead.

In response, Sam Altman stated that OpenAI's new defense contract includes protections against those same ethical red lines. Amodei, however, dismissed this in his memo, labeling OpenAI's messaging as "straight up lies" and accusing Altman of falsely "presenting himself as a peacemaker and dealmaker."

The central conflict stems from the DoD's requirement for AI to be available for "any lawful use," a clause OpenAI's blog post confirms is part of its agreement for "all lawful purposes." OpenAI clarified that it made explicit in its contract that unlawful uses, such as mass domestic surveillance, are not covered.

Critics argue this legal framework is insufficient, noting that laws can change and currently prohibited applications could become permissible. Public sentiment appears to align with this skepticism. Following OpenAI's DoD announcement, ChatGPT uninstalls surged by 295%.

Amodei noted this public reaction in his memo, writing, "I think this attempted spin/gaslighting is not working very well on the general public or the media, where people mostly see OpenAI’s deal with the DoW as sketchy or suspicious, and see us as the heroes." He added that while the narrative might influence "some Twitter morons," his primary concern is ensuring it does not resonate with OpenAI's own employees.