Government Contracts Update: Department of War/Anthropic Dispute and Downstream Implications for Contractors

The Department of War’s (“DoW”) dispute with Anthropic continues to evolve. Should you, as a government contractor, change your relationships with Anthropic or its products as a result? 

To recap the story:

• Two weeks ago, DoW designated Anthropic, a major provider of AI tools, including the only AI tool cleared for classified systems, as a supply chain risk following a breakdown in negotiations over contractual use restrictions on Anthropic’s Claude model. This marked the first time a U.S.-based company has ever been designated a supply chain risk. A few days prior, DoW considered using the Defense Production Act to ensure preferred access and potentially compel changes to Anthropic’s products.
• After the designation, the President instructed federal agencies to stop using Anthropic technology, and DoW statements implied contractors should not engage commercially with Anthropic.
• Multiple legal commentators questioned whether the Department’s actions exceeded contractual, legal, or regulatory authority.
• Last week, an eclectic group of leaders in business, tech, and other spaces, including many former government and military officials, urged the Senate and House Armed Services Committees to step in. They characterized the recent actions as “profound departure” from the law’s original purpose “to protect the United States from infiltration by foreign adversaries.”
• This week, Anthropic sued DoW and other agencies in the Northern District of California and the DC Circuit, alleging both legal and constitutional violations under the Administrative Procedures Act and the First and Fifth Amendments, respectively.
• The State Department has reportedly begun removing Claude from its systems.

 Immediate Steps for Government Contractors

Contractors using Claude or licensing other Anthropic products should have a clear idea of where those products exist in their systems. If you license Claude, you should review to determine whether you can exit and any risks inherent with continued performance. If you have licenses with multiple commercial AI tools (think: Microsoft’s CoPilot, OpenAI’s ChatGPT, or Google’s Gemini), you may get strategic about which to use in the short term. If you have not adopted AI into your business, you should do so thoughtfully, with an eye on functionality and evolving policy risk.

Looking Ahead

This administration’s rapid policy shifts continue. From efficiency to national security and trade, these shifts create actual and potential compliance issues for government contractors, even where indirectly involved. With AI, early assessments of contractual and supply‑chain risks should create better outcomes.

We are keeping an eye on the horizon for a few related developments: 

• Contractors and constitutional speech protections. While contractors mostly think of their contracts, we have seen renewed interest in broader legal themes based on certain performance-related requests, including constitutional concerns over contractors and speech. Can the government compel a software or AI company to alter its code or models to suit the government’s needs? Can the government compel a contractor to make political statements or take political positions relative to a contract? Should a contractor worry about terminations or other contract actions if it declines to make statements or take other speech-related actions? These open questions have resurfaced in this unusual environment. And now Anthropic has raised these issues in their court filings.
• The government contractor defense. The US Supreme Court recently issued an opinion affecting the government contractor defense, which ideally shields a contractor from liability when performing according to the terms of its contract. In short, the Court held that, while the defense may shield you from liability, it will not shield you from a lawsuit. How should you consider this when entering or performing government contracts?
• Resetting expectations around the FAR. The Revolutionary FAR Overhaul (“RFO”) kicked off by the administration last year has entered regulatory approvals. Some say it is not so revolutionary. We think, no matter how you interpret the changes, you should take this moment to re-examine your priors on the FAR. As long-time practitioners know, the FAR always evolves. Yet, with a one-time massive refresh, we should all ensure our intuitions honed over the years remain relevant and applicable. Sometimes that means going back to fundamentals.