The White House and Anthropic are closing in on an agreement that would put Claude inside the US intelligence community, according to The Information. The deal would clear the way for spy agencies to deploy Anthropic’s models on classified workloads, marking one of the most significant government AI contracts to date.
This is a big moment for Anthropic, and it tells you where the frontier AI race is heading next: national security.
What’s on the table
The Information reports that the arrangement would let US intelligence agencies tap Anthropic’s Claude models for their work. The specifics of scope, pricing, and which agencies are covered haven’t been fully disclosed, but the framing is clear: Claude is being lined up as an approved tool for the spy community.
For Anthropic, this puts the company on the same playing field as OpenAI, Palantir, and Microsoft, all of which have been racing to lock down federal and defense contracts over the past year.
Why this matters
Three things stand out:
- Anthropic’s positioning shifts. The company built its brand on safety-first AI research. A deal with US spy agencies signals it’s comfortable operating in the highest-stakes government environments, not just enterprise and developer markets.
- The government AI market is consolidating fast. OpenAI struck its own defense-adjacent deals earlier this year. Anthropic moving into intelligence work confirms that the top labs see government as a core revenue line, not a side bet.
- Classified deployments raise the bar. Running frontier models inside intelligence systems means new requirements around security clearances, isolated infrastructure, and auditability. Vendors that can meet those bars will dominate the next wave of federal contracts.
The bigger picture
A year ago, the conversation around AI in government was mostly theoretical. Pilots, task forces, executive orders. That’s changed. Anthropic already has an arrangement with the Department of Defense, and partnerships with Palantir and AWS to deliver Claude to defense and intelligence customers through secure cloud environments.
This new deal would deepen that footprint considerably. Intelligence agencies process massive volumes of unstructured data, signals, documents, intercepts, and that’s exactly the kind of work large language models are built for. Summarization, translation, pattern detection, drafting reports. If Claude gets approved for classified use at scale, it becomes embedded infrastructure inside the US government.
What to watch next
Expect three follow-ons:
- Pricing disclosure. Government contracts usually surface dollar figures eventually. The size will tell us how seriously agencies are betting on frontier models.
- Competitor response. OpenAI, Google, and Meta will push harder on their own intelligence and defense pitches. Watch for matching announcements within weeks.
- Policy debate. Civil liberties groups and Congress will weigh in. Anthropic has historically leaned into safety messaging, so how it frames this deal publicly will matter.
The race to put frontier AI inside the US national security apparatus is no longer hypothetical. It’s happening now, and Anthropic just moved closer to the center of it. Full details at The Information.