Last Updated on 22. June 2026
On the evening of Friday, June 12, 2026, the US AI company Anthropic received an export control directive from the Trump administration: access to its two newest models — Claude Fable 5 and Claude Mythos 5 — was suspended for all foreign nationals. Since a legally compliant separation between US and non-US users was not feasible on short notice, Anthropic disabled both models entirely — worldwide, for all customers, without warning. Three days after launch. This incident makes one thing clear:
Digital sovereignty in AI is not a political option — it is a strategic imperative.
The US government now treats cutting-edge AI technology like critical military hardware — with export bans, national security directives, and unilateral decisions that can cut off global users overnight. For Europe, this is a wake-up call that cannot be ignored.
Claude Fable and Mythos: What This Moment Means for Europe’s AI Sovereignty
We have been discussing digital sovereignty for years. We write policy papers, pass regulations, and launch initiatives. But as long as we depend on providers whose availability is determined by political decisions made in Washington for our business-critical AI infrastructure, sovereignty remains an empty promise.
What is new is the speed and directness with which technological dependency becomes operational vulnerability. Anyone running agentic AI systems built on US models today — for software development, process automation, or critical decision support — must ask: what happens when the switch gets flipped?
Digital Sovereignty in AI Means Freedom of Choice, Not Isolation
An important clarification: digital sovereignty in AI does not mean cutting ourselves off from powerful international technologies. It means retaining control over which technologies we use — and under what conditions.
In practical terms, this means companies and public institutions must be able to swap out AI models flexibly, keep sensitive data under their own control, operate agentic systems on-premises or in sovereign European cloud environments, avoid vendor lock-in, and keep the costs of their AI infrastructure permanently manageable.
This is not an ideological position. It is sound business sense.
ESACA: Europe’s Answer to AI Dependency
Together with Fsas Technologies, a Fujitsu company, mgm technology partners has launched the European Sovereign Agentic Coding Alliance (ESACA). ESACA is an open initiative to advance sovereign agentic software development in Europe. Our goal is to develop shared architectural principles, best practices, and reference solutions that enable organizations to deploy Agentic AI securely, economically, and independently — free from politically imposed access restrictions. Learn more
At the core of our work are sovereign software architectures, open standards and interoperability, cost-effective operation of agentic systems, security, governance and compliance, as well as the use of open and European technologies where they make sense.
At mgm, we bring decades of experience in model-driven software development, enterprise architectures, and the controlled use of AI. For us, ESACA is the logical consequence of what we see every day in client projects: the need for technological independence is real — and it is growing.
Act Now — Together
The challenges of digital sovereignty in AI cannot be solved by individual companies. They require collaboration between industry, academia, open-source communities, and public institutions. Europe has the knowledge, the talent, and the capacity for innovation. The Anthropic model shutdown on June 12, 2026 has shown: we also have the reason to act. What is missing now is the collective will to turn these strengths into sustainable, independent solutions.
We invite organizations that share this conviction to get actively involved. ESACA is open — to companies, research institutions, public bodies, and everyone who wants to help shape a sovereign European future for Agentic AI.
Building Sovereign Agentic AI for Europe.
Learn more and get involved.
Visit the ESACA website
