Anthropic Challenges the Pentagon: Battle Over AI Security

Anthropic Challenges the Pentagon on Risk Designation

Dario Amodei, the founder of Anthropic, announced on Thursday that the artificial intelligence company he leads is considering legal action against a recent decision by the U.S. Department of Defense (DOD). This decision, which classifies Anthropic as a risk to the supply chain, is deemed by Amodei to be "legally unfounded."

This announcement comes shortly after the DOD officially labeled Anthropic as a potential supply chain risk, following a weeks-long dispute over the level of control the military should have over AI systems. This status could prevent the company from collaborating with the Pentagon and its contractors. Amodei emphasized that the technologies developed by Anthropic are not intended for mass surveillance of American citizens or for creating autonomous weapons. However, the Pentagon insists on the need for unrestricted access for all legal uses.

Limited Impact on Clients

In his statement, Amodei highlighted that the majority of Anthropic's clients would not be affected by this designation. He explained that this measure primarily concerns the use of their AI model, Claude, in direct contracts with the Department of Defense, rather than all possible uses by other clients.

• "For our clients, this only applies to the use of Claude in direct contracts with the Department of Defense," he clarified.

Anthropic's Legal Arguments

Amodei provided an overview of the arguments that Anthropic might present in court. He pointed out that the DOD's letter, which classifies the company as a risk, is limited in scope.

• "This designation aims to protect the government rather than punish a supplier," he stated. Amodei added that the law requires the Secretary of Defense to use the least restrictive means necessary to ensure the protection of the supply chain.

Even for subcontractors of the Department of Defense, this designation should not restrict the uses of Claude or business relationships with Anthropic, as long as they are not directly related to specific contracts with the Department.

Internal Tensions and Memo Leak

Amodei also mentioned that constructive discussions had taken place with the DOD in recent days. However, these exchanges were disrupted by the leak of an internal memo. In this document, Amodei criticized OpenAI's relationships with the Department of Defense, calling them "security theater."

OpenAI recently signed an agreement to work with the DOD instead of Anthropic, which sparked internal reactions. Amodei apologized for the leak, asserting that the company did not intentionally share the memo.

Context and Outlook

The memo was drafted in a context of tension, shortly after several announcements, including a post on Truth Social by the president indicating that Anthropic would be removed from federal systems, followed by the risk designation for the supply chain by Secretary of Defense Pete Hegseth. Amodei described that day as "difficult" for the company and clarified that the memo did not reflect his well-considered opinions.

He concluded by stating that Anthropic is committed to ensuring that American soldiers and national security experts continue to have access to crucial tools, particularly in the context of operations in Iran. The company will continue to provide its models to the DOD at a "nominal cost" for as long as necessary.

Upcoming Legal Challenges

Anthropic may take this case to a federal court, likely in Washington. However, existing legislation makes this challenge complex, as it limits the usual avenues for appeal against government procurement decisions and grants the Pentagon broad discretion in matters of national security.

Dean Ball, a former White House advisor on AI, noted that courts are generally reluctant to challenge government decisions regarding national security, although it is not impossible.