The Pentagon Is Replacing Anthropic’s Claude — With OpenAI and Google

Two AI governance stories are converging this week — and together they reveal the most consequential regulatory divergence of the AI era.

The Pentagon Replacement

The Pentagon confirmed it is testing OpenAI and Google models to potentially replace Anthropic's Claude in classified systems.

This confirmation closes a loop that began in February 2026 — when the Trump administration blacklisted Anthropic after the company refused to let the Pentagon use its AI systems without safety guardrails for warfare applications. At the time, the Pentagon declared Anthropic a supply chain risk.

Google, Microsoft and Amazon were awarded classified network contracts instead.
Now the Pentagon is actively replacing Claude — which was previously the only AI model operating in its highest-classification networks — with OpenAI and Google alternatives. The companies that agreed to military deployment without safety conditions are taking the market share of the company that refused.
The commercial consequence of principle: Anthropic has lost its most strategically significant government customer.

The EU AI Act — 55 Days

The EU AI Act enforcement deadline moved to 55 days out.

Fifty-five days from today — August 4, 2026 — the EU AI Act's transparency enforcement provisions become binding. Every AI system generating synthetic content and deployed in Europe must be identifiable as AI-generated.

The watermarking requirement we have covered throughout our reporting arrives in less than two months.
For every GAFAM company operating in Europe — and for the Pentagon's AI deployments in NATO contexts — the August 4 deadline is now close enough to force decisions that have been deferred.

The Two Visions — Side by Side

The Pentagon replacement story and the EU AI Act countdown together create the clearest possible illustration of the divergence between American and European AI governance.

Washington's approach in 2026: AI safety is negotiable. Companies that accept military deployment without guardrails get government contracts. Companies that insist on safety conditions get blacklisted. The commercial incentive structure rewards compliance over principle.
Brussels' approach in 2026: AI transparency is mandatory. Companies deploying AI in Europe must disclose AI-generated content regardless of their business model or commercial incentives. The regulatory incentive structure rewards transparency over speed.

These two approaches are not merely different. They are structurally incompatible. An AI system optimised for US military deployment — designed to operate in classified networks without transparency mechanisms — cannot simultaneously satisfy EU AI Act transparency requirements for civilian deployment.
Every GAFAM company operating in both markets faces this incompatibility. Every day that Washington and Brussels move further apart on AI governance, the cost of operating in both markets increases.

What This Means for Anthropic

Anthropic's Pentagon exclusion and its WWDC Extensions inclusion are happening simultaneously. The company that lost its government AI contract because it insisted on safety guardrails has just gained access to one billion iPhone users because Apple chose it as a launch Extensions partner.

The market is simultaneously punishing Anthropic in the military sector and rewarding it in the consumer sector. The $30 billion funding round Anthropic closed in May — at a $965 billion valuation — suggests investors believe the consumer reward is worth more than the military penalty.

The European Perspective