All corrections
LessWrong February 25, 2026 at 11:05 PM

www.lesswrong.com/posts/rmYB4a7Pskw7DLpCh/anthropic-and-the-department-of-war

2 corrections found

1
Claim
Right now Anthropic is the only AI company whose models work in classified contexts.
Correction

This is incorrect: multiple other AI providers’ models have been approved for, or integrated into, U.S. government classified environments (including Google’s Gemini and OpenAI models via Azure OpenAI, and xAI’s Grok per DoD statements reported by major outlets).

Full reasoning

The post claims Anthropic is the only AI company whose models work in classified contexts. That “only” is contradicted by multiple independent sources:

  1. Google (Gemini) and xAI (Grok) in Pentagon networks (including classified):
    The Guardian reports Defense Secretary Pete Hegseth said the U.S. military would begin integrating xAI’s Grok into Pentagon networks, and quotes him referring to AI models being on “every unclassified and classified network.” The same Guardian piece also notes the Defense Department selected Google’s Gemini (in Dec. 2025) to power an internal military AI platform (GenAI.mil). If Gemini and Grok are being integrated/used inside Pentagon networks—including classified networks—then Anthropic is not the only AI company with models operating in classified contexts.

  2. OpenAI models in classified U.S. Government cloud environments (via Azure OpenAI):
    Microsoft’s official Azure Government blog states Azure OpenAI Service (which provides access to OpenAI models) was authorized by DISA for DoD Impact Level 6 (IL6) and that Azure OpenAI is authorized for workloads at all U.S. Government data classification levels. That directly contradicts the “only Anthropic” claim.

Because these sources document additional AI model deployments/authorizations in classified government contexts, the post’s “only AI company” statement is false.

2 sources
2
Claim
Only Claude is deployed on the classified networks used for actual sensitive operations, via Palantir.
Correction

This is contradicted by reporting that Grok and Gemini are being integrated/used within Pentagon networks, including for classified documents—so Claude is not the only model deployed in classified Pentagon environments.

Full reasoning

The post asserts that only Claude is deployed on classified networks used for sensitive operations.

However, TechCrunch reports that Defense Secretary Pete Hegseth said Grok would join Google’s Gemini in operating inside the Pentagon network, handling both classified and unclassified documents. That directly contradicts the “only Claude” claim: if Grok and Gemini are operating inside Pentagon networks and handling classified documents, Claude is not the sole deployed model in classified Pentagon environments.

Additionally, The Guardian similarly reports the Pentagon integrating Grok into its networks and references Gemini being selected to power a DoD internal platform, reinforcing that other models are deployed/being deployed in those environments.

2 sources
Model: OPENAI_GPT_5 Prompt: v1.6.0