Claude enters a classified intelligence environment
On 7 November 2024 Anthropic, Palantir and AWS announced that Claude 3 and 3.5 would be available on Palantir's platform inside an Impact Level 6 environment, accredited for data classified up to Secret.
Why it matters
The first time a frontier commercial model became available inside a classified enclave. Until then the line was whether a model could be sold to the military at all; from here on it ran through what the model was allowed to do once inside.
Palantir was the first partner to bring Claude models into classified environments; both Palantir and AWS hold Impact Level 6 accreditation. Palantir CTO Shyam Sankar: "Our partnership with Anthropic and AWS provides US defense and intelligence communities the tool chain they need to harness and deploy AI models securely." Named uses include foreign intelligence analysis and providing warning in advance of potential military activities. The source separately lists what Anthropic's usage policy barred at the time: disinformation, weapons design and use, censorship, malicious cyber operations, interference with military facilities, battlefield management and illegal weapons trafficking — but no blanket military prohibition of the kind Meta's policy carried. Palantir's own release was not read here: the investor page is client-rendered and returned only navigation in the browser pane. The record rests on a single secondary source.