In early March, the federal government told its own contractors that Anthropic was a security problem. The Pentagon labeled the company a supply-chain risk, a designation that barred it from working with other Defense Department vendors. The message was blunt: do not build on this company's models.
On June 29, the most populous state in the country did the opposite. Governor Gavin Newsom stood up and handed Anthropic every agency California has.
The deal makes Claude available to all California state agencies, plus its cities and counties, at a 50% discount, bundled with free workforce training and hands-on help from Anthropic's own engineers. It is the first time a single AI productivity tool has been cleared for procurement across the entire state government at once. And it lands as a pointed rebuttal to Washington, from a governor who has spent two years positioning California as the place that adopts AI on its own terms.
A Deal That Covers Every Agency
The agreement routes Claude through the California Department of Technology's new Statewide Information Technology Shared Services portal, known as SITeS, which centralizes approved AI tools with transparent pricing. State workers will use it for the unglamorous core of government work: drafting and summarizing documents, analyzing information, and answering the public faster.
Newsom framed the deal as a deliberate contrast with how AI is being adopted elsewhere. "AI should not replace the human work of government; it should help our workers move faster, solve problems more effectively, and deliver better results for Californians," he said in the announcement.
This is not a pilot. Several of California's largest agencies are already running Claude in production, and the new deal extends that access to every department that wants it.
| Agency or program | How it uses Claude |
|---|---|
| California DMV | Improving customer service and lowering wait times |
| Dept. of Health Care Services | Internal workflows to assist Medicaid recipients (the largest Medicaid agency in the country) |
| CDT and CalOES | Cyber defense: scanning, triaging, and patching state code with Claude Security and Claude Code |
| Engaged California | Powering a deliberative democracy platform for public input on AI policy |
| Poppy | A state-worker-built assistant with pre-built queries for common state tasks |
"As a California company, we feel a real responsibility to our home state," said Kate Jensen, Anthropic's Head of Americas. "We're honored to expand our partnership with California's agencies and to put Claude to work for the people who keep this state running."
California and Washington Split on Anthropic
The warmth of the California deal is impossible to read without the federal freeze-out sitting next to it.
Earlier this year, Anthropic and the Defense Department clashed over a contract that would have let the government deploy Claude for any lawful use. Anthropic wanted explicit carve-outs, language that would stop the government from using its technology to surveil Americans or run autonomous weapons without human oversight. Defense Secretary Pete Hegseth refused. The department signed with OpenAI instead, then went further and declared Anthropic a supply-chain risk, cutting it off from the broader Pentagon contractor ecosystem. Anthropic sued the administration over the blacklist.
The state insists the two tracks are unrelated. Chris Given, California's Chief Information Officer and the head of its Department of Technology, told POLITICO that the Pentagon's supply-chain-risk designation "just didn't come up" while the state negotiated its Anthropic contract. However convenient that sounds, it is also the whole point: California is buying on its own criteria, not Washington's.
For Anthropic, the strategic value is obvious. Shut out of the largest federal buyer, it just locked in the largest state one, in the state where 33 of the world's top 50 private AI companies are based, and where the company itself is headquartered.
What State Workers Actually Get
The deal is not only a discount. It comes with free training for the state workforce and technical support from Anthropic developers who help agencies design their own workflows. California has spent two years building toward this, developing more than 20 AI trainings for state workers and publishing guidelines on how agencies should evaluate AI's impact on vulnerable communities.
Some of the most consequential uses are already live. The Department of Health Care Services, which runs the largest Medicaid program in the United States, uses Claude in internal workflows meant to help caseworkers assist recipients. The DMV, the agency Californians love to hate for its lines, is using it to cut wait times. And in a detail that closes the loop on the Pentagon fight, two state bodies, the Department of Technology and the Office of Emergency Services, are using Claude Security and Claude Code for cyber defense, scanning and patching the state's own code. The company Washington called a security risk is being trusted to secure California's software.
The Other Side
A single vendor cleared to sell into every agency of the nation's largest state is exactly the kind of concentration that makes procurement watchdogs uneasy. "First cleared for statewide procurement" is a milestone for Anthropic and a dependency for California, one that will be hard to unwind if the relationship sours or the pricing changes after the introductory discount.
Civil-liberties advocates have a longer-standing worry: government AI touches benefits decisions, emergency response, and personal data, and errors there are not abstract. California's own March order exists precisely because AI in public services can encode bias or mishandle private information, and the state published equity guidelines warning about disparate impact on marginalized communities. The guardrails are real, but so is the risk they are meant to contain.
The sharpest caution comes from what happens when public-sector AI is trusted too far. A Utah AI chatbot recently recommended psychiatric medications without a doctor in the loop, a reminder that "productivity assistant" and "unsupervised decision-maker" are separated only by the discipline of the humans deploying it. Newsom's framing insists AI will support workers rather than replace their judgment. Whether that line holds across thousands of state employees and dozens of agencies is a question no press release can answer. The federal-versus-state fight over who gets to use which models sits inside a larger unsettled debate about state and federal authority over AI itself.
The Bottom Line
Two governments looked at the same company and reached opposite verdicts. The Pentagon saw a contractor that would not sign a blank check and cut it loose. California saw a homegrown company willing to work on the state's terms and handed it the keys to the government. Both decisions were about control: the federal one about what Anthropic refused to allow, the state one about what California insisted on getting.
For the hundreds of thousands of people who staff California's government, the practical result arrives quietly, as a new tool in the SITeS portal that drafts a memo or summarizes a case file. For the AI industry, the signal is louder. The most valuable customers in America are no longer only in Washington, and the criteria for winning them are diverging fast. Anthropic just proved a company can be blacklisted by the federal government and embraced by the largest state in the same season.
Given said the state simply went looking for the best tool at the best price. The Pentagon went looking for a contractor that would agree to anything. They found different companies. That is the whole story.
Sources
- Governor Newsom announces a first-of-its-kind partnership providing Anthropic tools to state agencies (Office of Governor Gavin Newsom, Jun 29, 2026)
- Anthropic and Gov. Newsom forge deal allowing California government to use Claude at half price (TechCrunch, Amanda Silberling, Jun 29, 2026)
- Exclusive: Newsom, Anthropic ink deal to expand government use (POLITICO, Jun 29, 2026)
- California Strikes Deal With Anthropic to Bring Claude AI to State Agencies (Decrypt, Jun 29, 2026)
- California signs deal to bring Claude AI tools to government workers (CBS Sacramento, Jun 29, 2026)
- It's official: the Pentagon has labeled Anthropic a supply-chain risk (TechCrunch, Mar 5, 2026)