Yesterday, the US government ordered Anthropic to stop providing its latest AI models – Mythos 5 and Fable 5 – to people without US citizenship. Anthropic then blocked access completely. The explanation: at present, the company cannot reliably determine which users hold which citizenship.
Let that sink in for a moment. Access to a key technology was withdrawn overnight by executive order. No transition period, no choice for the affected companies and public administrations in Europe. To this day, the US government has not provided further details on its security concerns. But for the real lesson, the precise justification is secondary. What matters is the fact that access to critical technology can be taken away from us at any time and without a comprehensible explanation.
Markus Beckedahl from the Center for Digital Rights and Democracy summed it up in an interview with Tagesschau: this is a precedent. It is being enforced through so-called export controls – in the name of a vague notion of “national security.” It affects not only European companies, but even foreign nationals living in the United States, including Anthropic employees without a US passport. We should prepare for the possibility that our entire infrastructure based on American software and hardware could be switched off at any time. That means we need alternatives now in order to become digitally sovereign.
This is about more than mere dependency
Dependency alone is already reason enough to be alarmed. But there is a second aspect that is at least as important: some of the leading AI providers are working with the US government to advocate very strict AI regulation, thereby deliberately making market access more difficult for competitors, especially from the open source ecosystem. This is justified by pointing to the risks of the technology. In practice, however, it is also – at the very least – about securing a competitive advantage and geopolitical dominance in AI.
Even voices close to Silicon Valley read the situation clearly. On the All-In Podcast, David Sacks called it a “regulatory capture campaign based on fear-mongering” – a strategy of regulatory capture built on fear. The mechanism behind it is simple: create fear about the dangers of frontier AI, use that fear to call for strict government regulation, and ultimately benefit from the fact that large, well-capitalized labs can meet such requirements while smaller competitors and open source developers cannot.
The risks of AI are real. The decisive question is not whether safeguards are needed, but where they should apply. If they are placed at the very top, at the level of model access, a small number of licensed labs decide who gains access to knowledge and capabilities. If they are applied further downstream, for example through liability for actual misuse, innovation remains broad-based and verifiable.
Why open source is the real point of contention
The concern that large closed labs have about open source AI is justified – but not for the reasons they publicly state. Open source models are improving rapidly, are significantly cheaper and are coming ever closer to the performance of closed models. That is precisely why they create direct competitive pressure. At the same time, they are also the escape route for all those who cannot work under the restrictions of closed models.
Open source is therefore not a side issue in this debate. If frontier AI is regulated like aviation or pharmaceuticals, a small number of licensed, audited and well-capitalized labs will dominate in the end. If open models remain freely available and continue to improve, innovation, sovereignty and local control will be preserved.
Europe must respond now
Yesterday should be a wake-up call for Europe. We cannot stand by while our public administrations, hospitals, companies and research institutions build on an infrastructure that can be taken away from us with the stroke of a pen. Beckedahl identifies the danger clearly: this is an instrument of power, a means of coercion – and we are exposed to it if we continue to make ourselves completely dependent on a small number of companies headquartered in the United States.
The European Commission has recognized the problem in principle and launched the EU Tech Sovereignty Package on June 3, 2026. But a package on paper is not enough. It remains to be seen whether we are prepared to invest serious money in alternatives based on open standards and designed with encryption in mind from the outset.
This is precisely where open source AI comes in – and precisely why we must build on it now. This is not about closing off markets. It is about sustainably securing this technology for everyone, including Europe. It is about use and cooperation on equal terms, and about full transparency as protection against unwanted influence over public opinion.
The Open Source Business Alliance also warned explicitly today against exactly this pattern: market dominance is used to suddenly and uncompromisingly block access. What began with Microsoft and cloud services is now continuing in AI. Open AI models based on open source cannot be switched off. They offer stable structures and genuine competition that benefits the European software industry. For this to succeed, European providers of digitally sovereign and sustainable AI need support in competing with US providers. Then Europe can catch up and become more innovative itself.
What Europe can do in concrete terms
Europe should use open source AI exclusively in administrative procedures. The state must not bind its core processes to providers that can lock it out at any time.
Europe should create incentives and strengthen competitiveness. One conceivable model would be this: European data centers that offer open source AI while meeting sovereignty and energy sustainability criteria could receive reductions in energy taxes or other benefits that can be passed directly on to prices.
And Europe should directly support open source AI for a transitional period – but in a way that preserves genuine competition between different providers.
Yesterday’s shutdown was an instructive lesson. And it has long since ceased to be only about AI. It is about our entire digital infrastructure. Open source is not Plan B. Open source is the only path that secures sovereignty, transparency and innovation at the same time.
