Anthropic CEO Dario Amodei announced on Thursday that he cannot comply with the Pentagon’s request for unrestricted access to its AI systems, citing ethical concerns. He stated that while the Department of War makes military decisions, certain AI applications could threaten democratic values and exceed the current capabilities of technology.
The specific concerns include mass surveillance of Americans and the deployment of fully autonomous weapons without human involvement. The Pentagon believes it should have full access to Anthropic’s AI for all lawful purposes and contends that a private company should not dictate its usage.
This announcement precedes a Friday deadline set by Defense Secretary Pete Hegseth for Anthropic to comply or face potential labeling as a national security risk or be subjected to mandates under the Defense Production Act (DPA), which could force the company to prioritize defense production.
Amodei highlighted the inconsistency in these threats, noting the paradox of being labeled both a security risk and an essential security asset. He urged the Department to reconsider its stance given the value Anthropic’s technology offers to the military but acknowledged the Department’s right to select its contractors.
Anthropic is currently the only AI lab with classified-ready systems, though the Department of Defense is preparing xAI for similar roles.
Amodei expressed a preference to continue collaboration with safeguards but stated that if the Department chooses otherwise, Anthropic will aid in transitioning to another provider to ensure military operations remain unaffected. In summary, Amodei suggested parting ways without conflict if necessary.