In his recent letter to employees, Anthropic CEO claimed that the Department of Defense wanted them to delete a specific phrase preventing the exact type of mass surveillance Anthropic was concerned about.

In a recent internal letter to employees, Anthropic CEO Dario Amodei revealed that the U.S. Department of Defense pressured the AI safety company to delete a specific constitutional clause from Claude's operational framework. The clause in question explicitly prevented the AI system from being used for mass surveillance operations—a core concern in Anthropic's AI safety principles. This disclosure comes amid growing scrutiny of government-AI industry relationships and follows Anthropic's $2-4 billion valuation and partnerships with major tech firms.

According to the letter, the DoD's request specifically targeted language in Claude's constitution that would have prohibited its use in broad surveillance applications. While Anthropic's response to the pressure remains unclear, the incident exposes the tension between AI safety commitments and national security demands. The revelation raises significant questions about transparency in AI development and whether other AI companies face similar pressures from government agencies seeking to bypass ethical safeguards for security purposes.