Dario Amodei's Statement on Discussions with the U.S. Department of War

Anthropic emphasized the importance of using AI to defend the United States and democratic nations, stating that it has already been actively deploying its models to the Department of War and intelligence agencies. It has expanded to classified networks, national laboratories, and custom models for national security customers, but drew the line at two use cases.

• Large-scale domestic surveillance: While it supports lawful overseas intelligence and counterintelligence missions, it views large-scale surveillance of U.S. citizens as incompatible with democratic values.
• Fully autonomous weapons: Current frontier AI lacks sufficient reliability, so it cannot be provided for weapons that select and engage targets without human control.

The company noted that it has already built these conditions into its contracts, and stated firmly that it cannot agree to the Department of War's demand to remove them. It reported that the Department of War is demanding acceptance of "any lawful use" and the removal of safeguards, and has raised the possibility of removing its systems, designating it a supply chain risk, and even invoking the Defense Production Act if it refuses.

Anthropic stated that while a company cannot make the Department of War's military decisions for it, given the significant value its technology provides to the military, it wants to continue cooperation while maintaining its safeguards. It added that even if the Department of War switches suppliers, it will support the transition to another provider so that military planning and operations are not disrupted.