Europe cannot continue building its tech stack on access that can be switched off overnight by a foreign government. The US abruptly shut down access to Anthropic’s most powerful models, Mythos and Fable 5, by non-US citizens on June 12 over national security concerns. What’s become clear is that the US has a “kill-switch” on AI services that the UK and Europe depend on, making sovereign AI infrastructure more urgent than ever.
The anxiety about the world’s dependence on US and Chinese tech giants has escalated. If Trump can pull the plug now, when countries are still piloting AI solutions – what will happen when hospitals, militaries, police forces and even governments have transformed their operations to become totally dependent on US AI?
This wake-up call has shown that AI is not just a tool but critical infrastructure. It’s as essential to societal functioning as the internet, oil or electricity – and we’ve all witnessed the havoc that broke out when Europe’s biggest power cut hit Spain last year, or when oil tankers were blocked from the Strait of Hormuz.
Sovereignty means control, and the location of the headquarters matters less than your ability to york or move your software. It’s imperative now that European countries move beyond pilots to build their own robust sovereign AI infrastructure that prioritises privacy.
[ 👋 Hi there! If you’re here to find out more about FLock.io, follow us on X and email us at hello@flock.io to learn about how we help companies and governmental organisations build sovereign AI solutions.]
In tech, the infrastructure layer is whatever layer becomes the standard that everyone else builds upon. Once, a company’s backbone was the facilities. With digital transformation, it was IT. In 2026, renting AI like SaaS is a massive pitfall as AI becomes the foundation of business operations. Tomorrow’s medical diagnostics, surgeries and survival rates will rely on AI-driven systems. The future of warfare will hinge on who has the most capable model used in drones, counter-drone defence systems and cybersecurity.
At FLock.io, we firmly believe that compliance is non-negotiable, and that sensitive data staying local should be a no-brainer. But this kill-switch incident has escalated the urgency of sovereign AI to protect a nation’s infrastructure from being unplugged with the flick of a switch. In this blog, we explain how Europe can act.
The EU has finally woken up to the risks
A sovereign AI infrastructure stack refers to a system where data, compute, and AI processing are governed entirely within its jurisdiction. But it’s a distant dream unless Europe develops and owns its own privacy-preserving infrastructure; and a model’s guardrails are futile when what truly matters is who holds the reins.
Brussels is aware of this danger. So much so, that it seems sovereign AI could be Europe’s next industrial revolution. Innovating, investing in and adopting it could powerfully reignite labour productivity growth. This time, it’s not just about performance but jurisdiction.
The European Parliament officially ditched Google over privacy concerns in June, replacing it with Qwant, a privacy-first search engine from France. The decision was made “in line with the Parliament’s commitment to digital sovereignty and the protection of users’ personal data”. Further afield, several countries (Australia, South Korea, Taiwan, and others) restricted or banned the Chinese AI company DeepSeek’s products on government devices back in 2025.
The European Commission also unveiled its Tech Sovereignty Package this month: a set of legislative and strategic measures designed to reduce the EU’s reliance on foreign technology in favour of local alternatives.
It’s a step in the right direction. The goal is to elevate open-source software from a research tool to a structural lever of sovereignty. It aims to triple Europe’s data centre capacity over the next five to seven years so data doesn’t have to be sent to other continents. It will also map out just how heavily public and private infrastructure relies on non-EU providers.
Corporations, too, are beginning to mix European, US and Chinese models.
Siemens, Renault, Orange and ChapsVision have said they already mix US, Chinese and European models. Orange said the Anthropic restrictions made it “patently clear, if it wasn't before, how important it is for Europe to have access to an AI service that it can control, that will never be switched off on a whim.”
Europe has a large tech gap to fill
France and Germany are the countries predominantly spearheading a continent-wide push to achieve digital independence from US and Chinese tech giants. But while the EU’s Tech Sovereignty Package is a step in the right direction, there’s a large tech gap to fill.
Europe lacks a frontier AI company (Mistral AI in Paris is the strongest contender but is a shadow of OpenAI’s size). We are seeing a growing ecosystem of foundation and open-source model developers, as well as applaudably responsible data governance and privacy-centric model development. However, there is limited access to large-scale compute and high-quality multilingual datasets.
Only 20 per cent of the EU’s tech is from inside the EU, and together the US and China control an enormous 90 per cent of the world’s AI computing infrastructure. US companies have also invested about five times more in AI than their European peers, according to a report by McKinsey & Company. When considering geopolitical tension, Europe’s reliance on foreign tech becomes a critical vulnerability.
The window opportunity for Europe to capture the value and set the standards (regarding privacy, technicality and ethics) is narrow but clear. The continent is home to some of the highest ranking academic institutions. Seizing this chance demands urgency, unity and scale from the private and public sectors.
If Europe sleeps on it, its competitive position will be worsened at a time of declining economic growth. If it steams ahead, a high-adoption and high-sovereignty AI scenario could increase Europe’s annual GDP by €480bn by 2030, compared to €375bn in a high-adoption, low-sovereignty scenario, or €80bn for an entirely missed opportunity, according to McKinsey.
Even setting aside the kill-switch risk for a moment, centralised AI is inherently incompatible with governmental organisations.
The biggest impediments to Europe adopting AI are concerns over security, sovereignty and trust
The biggest obstacle to adopting AI across the continent are concerns over security and trust. By addressing these through instilling privacy and safety into the architecture, Europe can overcome this.
A recent McKinsey survey of European companies found security and sovereignty topics were a key driving factor behind lagging adoption of cloud and AI service solutions. Forty-four per cent of tech leaders cited concerns about data security as a reason for not using the public cloud, while 31 per cent said the need to store data in a specific country or region prevented them from doing so. European options were the preferred for hosting.
Safety doesn’t have to compromise competitiveness or collaboration
When the US severed access to Fable 5, British hospitals were suddenly stripped of access. In retaliation, MPs criticised the UK for prioritising safety over building competitive AI capacity at a time when we desperately need to wean off Silicon Valley. But actually, safety and competitiveness are not mutually exclusive. Quite the contrary – done right, prioritising sovereignty can mean more robust infrastructure and improved collaboration between nations and organisations.
Federated learning (FL) has emerged as the future of privacy-preserving AI training. It was invented in 2016 as a way for multiple parties to improve a model with enhanced data security. FL is how governmental organisations, financial services and the medical sector are training models on local hardware. It’s a GDPR-compliant alternative to traditional machine learning. It means never surrendering data to foreign corporations, and a foreign country can’t turn off their entire operations.
Sarawak, Malaysia was the proof of concept for FL-enabled sovereign AI
FLock.io is the key technical partner in a government initiative to develop sovereign AI for Sarawak, the largest state in Malaysia. This will be a touchstone for European countries like Greece, Estonia and Malta who are pursuing sovereign AI.
Sarawak aims to become a fully AI-driven state within the next five years. Doing it right means preserving data sovereignty, supporting local contexts, and avoiding dependence on foreign data centres. FLock.io worked with the Sarawak AI Centre to demonstrate how this can be achieved in practice using FL Alliance, FLock’s federated learning framework. The experiment showed that distributed inference enables a large model to run efficiently on smaller, local GPUs in e.g. a hospital instead of relying solely on centralised data centres.
Sovereign AI is a big investment – choose it if compliance is non-negotiable
Choosing to build a sovereign AI stack is not a small investment. If you are a regulated organisation in sectors including banking, healthcare, government or defence, you handle sensitive personal data at scale, or your priority is compliance and full legal control.
The layers with the greatest data sovereignty risk, which hold the most sensitive information, should be prioritised first. Long-term autonomy requires that sovereignty reaches further down the stack, but innovating at the application and models layers is the first step. Models must be built that are attuned to a European context and all the languages and regulations that come with it.
More about FLock.io
FLock.io is an AI research and infrastructure company pioneering enterprise-grade federated learning and distributed AI solutions. Its decentralised federated learning architecture and production-ready platforms (AI Arena, FL Alliance, and FLock API Platform) enable organisations to train and deploy their own custom AI models on local hardware while maintaining full data privacy, model ownership, and regulatory alignment by design.
[ 👋 Hi there! If you’re here to find out more about FLock.io, follow us on X and email us at hello@flock.io to learn about how we help companies and governmental organisations build sovereign AI solutions.]
