Why cutting Anthropic from government could be harder than expected

The Trump administration will head to court this week to defend its decision to ban the federal government from using Anthropic products, but removing the technology itself may prove to be the bigger battle. 

The public clash between the Pentagon and Anthropic is forcing the federal government to confront just how deeply embedded AI has become in Washington — from the top of the chain in agencies down to private contractors. 

Like most AI firms, Anthropic’s presence in government work runs deeper than just contracts. Even as tech firms compete, they also routinely do business with each other, and industry observers warn that entanglement complicates efforts to scrub the technology from all government systems. 

“There are multiple reasons why these apparent competitors might have an interest in mutually supporting each other, because it is part of one big ecosystem,” said Sarah Kreps, the director of the Tech Policy Institute in the Cornell Brooks School of Public Policy.

“If one company fails, then there is a way in which the trust of the entire enterprise could be imperiled,” Kreps added. 

In an unprecedented move, the Pentagon designated Anthropic a supply chain risk earlier this month, restricting the military and defense contractors from using the company’s product. President Trump also directed federal civilian agencies to “immediately cease” using the technology via a social media post. 

Anthropic sued the Trump administration, asking a federal court in California to issue a temporary halt to the supply chain risk designation and Trump’s directive. A judge will hear from both sides Tuesday in a hearing over the request. 

The contract at the heart of the Pentagon clash reflects the interdependent tech ecosystem the Trump administration finds itself trying to unwind by removing Anthropic. Nearly two years ago, Anthropic struck a partnership with longtime government contractor Palantir to host its AI models in government agencies, allowing the AI firm to skip the standard, lengthy security authorization process, known as FedRAMP. 

Palantir also has Anthropic integrated into its own systems, which are used across the government. CEO Alex Karp told CNBC earlier this month that the company plans to add other models amid the dispute, but that it is still using Anthropic.  

Claude is also offered on the Copilot platform hosted by Microsoft, one of the largest IT contractors for the government. Meanwhile, Google, which has its own series of AI models called Gemini, has also invested billions into Anthropic, owning at least 14 percent of the AI firm, while Amazon invested $8 billion into the company and hosts Anthropic in its Amazon Bedrock platform. 

Further, Anthropic has emerged as a standout AI tool for its abilities, including Claude Code, a coding tool that helps people or companies build features, fix bugs and automate tasks. 

“Claude Code, in tech circles, [is] all that people have been talking about for months now,” said Michael Boyce, former director of the Department of Homeland Security’s AI Corps program. “It’s an amazing tool. While there are other strong competitors in the space, it continues to be field defining.”

“What Claude does is different from what any other platform or any other system has been able to do,” Kreps added, stating the available alternatives are “not as good.” 

Anthropic’s request has received support from large swaths of the technology industry, which have their own products and interests to consider, too. Confusion remains over just how far down the designation will stretch.

“Companies are stepping up because they don’t want to be next,” Franklin Turner, the co-chair of McCarter & English’s Government Contracts practice group, told The Hill. “To allow this kind of thing to go unchallenged, I think a lot of folks believe [it] would be probably irresponsible from a corporate standpoint.” 

Microsoft was one of the first tech firms to publicly back Anthropic’s request to temporarily halt the supply chain risk designation. A spokesperson for the company said “everyone involved shares common goals” and time is needed to “find common ground.”

Other technology companies, think tanks and trade associations followed Microsoft’s lead, submitting briefs in the California case, along with Anthropic’s separate suit in the D.C. court of appeals.

The App Association, a global trade association for small and mid-sized tech companies, cited various scenarios of concern raised by its members in its amicus brief.

In one case, the association said a two-person startup selling logistics software to a Defense Department prime contractor used Claude Code to write its entire testing software, but it is “functionally indistinguishable from hand-written code.” 

“It may be plausible for a small developer contracted to work with the Department of War to abstain from using Claude in its own processes, but is quite implausible for that developer to know whether any of the tools it uses were coded by others using Claude,” the association wrote, using the Trump administration’s preferred name for the Department of Defense.

In another brief, dozens of workers with OpenAI and Google, including Google chief scientist Jeff Dean, argued the U.S.’s “thriving AI ecosystem leads the rest of the world largely due to the competition and flow of ideas between different AI companies.” 

Some of this support has come through back-channel discussions, The New York Times reported, citing numerous unnamed current and former technology company employees. 

Emil Michael, undersecretary of Defense for research and engineering, acknowledged the complexities of eliminating Anthropic as a vendor from government systems. 

“You can’t just rip out a system that’s deeply embedded overnight,” Michael told CNBC. 

The Pentagon set a 180-day deadline to remove all of Anthropic’s AI products from its systems and directed any other company with business ties to the Defense Department to halt using any of the firm’s products on work tied to defense contracts in the same timeline. 

Turner predicted companies will face “herculean hurdles” for the “supply chain cleansing exercise.” 

“The upshot is it’s going to be very hard for most contractors to certify that Anthropic is nowhere in their supply chain,” Turner said, adding Anthropic “produces all sorts of things, and can be used to produce all sorts of things, including open source code, algorithmic open source code.” 

Pentagon chief information officer Kirsten Davies said she will grant exceptions for “mission-critical activities directly supporting national security operations where no viable alternative exists,” but it is not clear how flexible this will be. 

Some civilian agencies are already tackling the removal, including the Department of Health and Human Services (HHS), which was one of the first to disable Claude for thousands of employees. 

In a notice to agency AI leaders obtained by The Hill, the HHS’s Office of the Chief Information Officer asked staff to list its current or planned use of Claude, including whether contractors or subcontractors use the models and if vendor systems embed the technology. 

“In every agency, there is some vendor being used, or subcontractor that is using a service that has an Anthropic model somewhere,” one HHS leader told The Hill on the condition of anonymity.

Copyright 2026 Nexstar Media Inc. All rights reserved. This material may not be published, broadcast, rewritten, or redistributed.

Senate GOP says Trump signals shift to backing DHS compromise with Democrats

US issues ‘worldwide caution’ alert to travelers

Van Hollen: Trump is ‘lying’ about talks with Iranians

GOP cracks in Senate begin to show in DHS shutdown fight

USS Gerald R. Ford aircraft carrier leaves Middle East amid Iran fight

Democrats release details behind spending on Noem’s $220M ad campaign

Pentagon closes office space for journalists after judge’s ruling on building ...

Homan: ICE officers will not assist with airport security scanning amid TSA ...

Pentagon Press Association calls for immediate restoration of ...

Bannon says ICE at airports ‘test run’ for 2026 elections

Homan expects ICE protests at airports, sees DHS shutdown ending soon

Iran war takes mounting toll on America’s military

Kentucky’s Andy Beshear takes aim at Vance as new rivalry intensifies

Democrats’ quest for relatable white dudes finds new candidates

Netanyahu says Trump sees opportunity for deal with Iran

Key things to watch for in Florida special election on Trump’s turf 

Trump: US isn’t negotiating with Iran’s supreme leader

Iran targeting buyers of US Treasury bonds: Parliament speaker

© The Hill