Microsoft sticks with Anthropic despite Pentagon blacklist

The Pentagon just slapped Anthropic with a supply-chain risk label, but Microsoft isn't backing down.

While some defense tech contractors rushed to scrub Claude from their systems, Microsoft became the first major tech giant to push back, announcing on Thursday that it will keep Anthropic's models humming across its flagship enterprise platforms.

According to an internal legal review, Microsoft concluded that the Pentagon's designation only applies to the Department of War itself. That means developers and corporate clients using Microsoft 365 Copilot, GitHub Copilot, and Azure AI Foundry can continue using Claude as usual.

A messy breakup in Washington

The clash follows a bitter falling out between Anthropic and defense officials over mass domestic surveillance and fully autonomous weapons.

After talks collapsed, President Trump called on federal agencies to ditch Anthropic, and Secretary of War Pete Hegseth gave the company a six-month deadline to wind down its Pentagon operations. The irony? CNBC confirmed Anthropic's models were actively used in recent U.S. airstrikes against Iran.

Anthropic's chief rival, OpenAI, wasted no time filling the void, quickly agreeing to run its own models for the Pentagon's classified workloads. Anthropic, meanwhile, plans to fight the supply-chain designation in federal court.

Billions at stake in the cloud

Microsoft's defense of Anthropic is fundamentally rooted in cold, hard cash and enterprise software reality.

Anthropic has committed a massive $30 billion toward Microsoft Azure cloud spending, while Microsoft pledged up to $5 billion in the startup. While smaller than Microsoft's historic relationship with OpenAI, it remains a pillar of CEO Satya Nadella's "model choice" playbook, which gives business users flexibility to switch between rival models.

Software engineers have adopted Claude heavily for drafting code. Tearing Anthropic out of GitHub Copilot would have caused massive friction for paying enterprise accounts.

Why it matters

For everyday developers and enterprise teams, this provides immediate breathing room. If you rely on Claude for coding in GitHub Copilot or drafting documents in Office, your workflow is safe for now, provided you aren't doing work directly for the Department of War.

For the AI landscape, Microsoft just drew an indispensable firewall. The Pentagon tried to use supply-chain risk powers to ostracize a leading lab over ethical pushback. If Microsoft had folded, government contracts would have dictated civilian enterprise software overnight.

We think Microsoft played this masterfully. It complied with the letter of the law by cordoning off the military, while protecting its $30 billion cloud deal and preserving its reputation as an open, multi-model platform.

However, this balance looks precarious. As frontier AI labs increasingly split between those willing to do unconstrained defense work and those setting strict red lines, enterprise tech giants will find it increasingly tricky to straddle both worlds.