Pentagon Says It Has Stopped Using Claude As Meta And Microsoft Scale Back Internal Use
· TexaHub · AI in Industry, Geopolitics & Tech, LLMs & Foundation Models
The US Department of Defense told the BBC it has ceased using Anthropic products, while The Information reported that Meta and Microsoft are cutting internal Claude use. The reports come weeks before Anthropic's planned IPO.
Pentagon confirms phase-out
The Pentagon has told the BBC that it "has ceased the use of Anthropic products," completing a phase-out that followed the department's designation of Anthropic as a national security supply-chain risk, according to reports on 6 October 2026 [1][4]. The dispute began when Anthropic CEO Dario Amodei refused an "any lawful use" contract term, objecting to mass domestic surveillance and fully autonomous weapons uses [3].
BBC sources said, however, that Claude was still in use the previous week for military operations against Iran through Palantir's Maven Smart System, which would conflict with the official statement. Anthropic declined to comment on the discrepancy, according to AI Weekly [4].
Why removal took months
Claude had been integrated into Maven, the platform used to process drone footage and satellite imagery. Lauren Kahn of Georgetown's Center for Security and Emerging Technology said that once such models are integrated "it can be painful to remove them" [3]. The Pentagon has since moved to models from Google, xAI and OpenAI, according to the same reports [3][4].
On 25 September the D.C. Circuit rejected Anthropic's challenge to the supply-chain-risk designation, finding that use restrictions could legitimately factor into procurement decisions, according to Runtime Wire [3]. Legal proceedings between the company and the Pentagon are described as ongoing [1].
Meta and Microsoft cut internal usage
Separately, The Information reported on 6 October that Meta and Microsoft are reducing internal use of Claude as they promote their own tools, according to summaries of the report [1][2]. Meta's Claude Code users fell from about 60,000 to about 30,000, while its in-house MetaCode has more than 30,000 users and Muse Code more than 6,000. Meta spent over $105 million on Claude Code in one recent 28-day period [2].
Microsoft's spending on internal Anthropic use, once on pace for at least $1 billion a year, has fallen, and work is shifting to GitHub Copilot CLI and OpenAI models, according to CryptoBriefing's summary [2]. Heise reported the decline as roughly one-third [1]; the two outlets' figures differ in detail.
IPO context
Anthropic's IPO filing disclosed that two unnamed customers account for about 25 percent of revenue, which CryptoBriefing said suggests internal-usage arrangements can be fragile. Growth in customer-facing Claude usage through Azure and Amazon Bedrock partly offsets the losses, it added [2].
Sources
[1] Anthropic: Pentagon completes phase-out, Meta and Microsoft restrict usage — heise online, 6 Oct 2026
[2] Meta and Microsoft cut back on Anthropic's Claude as in-house AI tools mature — CryptoBriefing, 6 Oct 2026
[3] Pentagon says it has stopped using Anthropic's Claude after a six-month phaseout — Runtime Wire, 6 Oct 2026
[4] Pentagon tells BBC it has stopped using Anthropic's Claude — AI Weekly, 6 Oct 2026