European countries to watch in sovereign AI : France, Germany, Estonia and Greece
The US curbing the world’s access to two Anthropic models has been a wake-up call: without sovereign AI, Europe could be switched off overnight. It has spurred the continent to seek domestic alternatives, making 2026 the most pivotal year yet in the developing AI that the bloc can control.
France and Germany are indisputably at the forefront of the sovereign AI push. But among the most interesting countries to watch are Estonia and Greece. The UK, however, sits at a different crossroads. This blog explains why this is, and how FLock.io is leveraging privacy-by-design federated learning to help governmental agencies build their own models while keeping their data secure.
[ 👋 Hi there! If you’re here to find out more about FLock.io, follow us on LinkedIn and X and email us at hello@flock.io to learn how we help companies and governmental organisations build sovereign AI solutions.]
Europe is finally getting real about sovereign AI
The European Commission unveiled its Tech Sovereignty Package on June 3, 2026 (dubbed “Tech Liberation Day”): a set of measures designed to reduce the EU’s reliance on foreign technology ​in favour of local alternatives.
The EU has opened 19 “AI Factories”: publicly funded ecosystems built around AI-optimised supercomputers to provide startups, SMEs and researchers with free computing power. They are in cities including Athens, Groningen, Barcelona, Bologna, Ljubljana and Kajaani.
However, the continent relies on China for hardware and raw materials, and on the US for AI software and cloud infrastructure. Europe controls less than 5 per cent of global frontier-scale AI compute. Europe also lacks a frontier AI company (Mistral AI in Paris is the strongest contender but is a shadow of OpenAI’s size).
But the reality is that you don’t need a massive, expensive corporate cloud to build custom AI.
Federated learning has emerged as privacy-by-design technology
Given the lack of cloud infrastructure in Europe, federated learning (FL) has emerged as a clear way forward. Instead of data centres (which carry environmental downsides), FL is a greener and more secure way to train models using local hardware.
Instead of moving large, sensitive datasets to a central server, the algorithm travels to the local data source (e.g., hospital, bank, or device), trains and sends back only aggregated model updates. FLock.io provides FL to be a GDPR-compliant alternative to traditional machine learning. It means never surrendering data to corporations, and a foreign country can’t turn off their entire operations.
France and Germany are the indisputable frontrunners in sovereign AI
France and Germany are leading the charge in sovereign AI: the push to build and control AI within Europe’s own borders. They want to protect citizen data, prevent reliance on US tech giants and ensure AI reflects European values.
France is moving to cut its ties with Palantir, replacing it with the French data analytics firm ChapsVision. The country’s domestic intelligence agency began to use the controversial US data platform as a temporary solution after the 2015 terrorist attacks on Paris, but has long described it as a temporary solution until a domestic alternative became available. France also plans to launch a conversational AI assistant for government employees using tech from French startup Mistal AI to protect state secrets.
With the launch of the Deutschland-Stack, and with the help of Deutsche Telekom, Germany is building a sovereign AI platform for the federal government. The first uses will be document processing, knowledge management and text summarisation.
Both France’s and Germany’s national institutes for AI also announced a binational AI centre together. Τhe two countries also co-authored a joint paper on the urgency of digital sovereignty on June 17.
Outside of Europe, FLock.io demonstrated FL for sovereign AI in Southeast Asia
FLock.io is the key technical partner in a government initiative to develop sovereign AI for Sarawak, the largest state in Malaysia, including for healthcare and the civil service. This will be a touchstone for European countries that are pursuing sovereign AI but do not want to send data to a multi-billion-euro corporate data centre with potentially disastrous environmental consequences.
The result of a project with the Sarawak AI Centre was a world-first achievement: FLock.io fine-tuned an LLM using FL with an indigenous language. This is a milestone for underrepresented languages at a time with standard models usually default to replying in English.
It showed that distributed inference enables a large model to run efficiently on smaller GPUs instead of relying solely on centralised data centres, which pose greater environmental challenges. This more sustainable technique involves sharding large models and deploying them across smaller, distributed hardware infrastructure, such as networked university labs and office computers. Researchers in trained a custom AI model using a local network of 25 consumer-grade graphics cards (NVIDIA RTX 5090s) housed entirely inside a university campus.
The platform will subsequently be deployed by hospital partners in the US, Europe and China. In 12 months or more, it will target a gap of approximately 30 countries in domestic AI compute. It will establish a standard for cross-border healthcare AI collaboration in the Asia-Pacific and Europe.
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.
Greece has one of the world’s most powerful supercomputers
Greece’s new national supercomputer DAEDALUS just entered the global supercomputer rankings, strengthening European technological autonomy. The system ranked 31st in the TOP500 list of the world’s most powerful supercomputers and 23rd in the Green500 list of energy-efficient systems.
Greece, the cradle of democracy, seeks to put humanity ahead of AI by adding safeguards to its constitution. On May 7, 2026, Prime Minister Kyriakos Mitsotakis outlined the proposed amendment to require that AI serve human society with the aim of protecting future generations. The proposed changes would include a provision that says: “AI shall serve the freedom of the individual and the prosperity of society, ensuring that risks are mitigated and that the advantages it provides are fully realised.”
The country is transitioning from strategic planning to the early implementation of its national AI agenda. It’s currently building the Pharos AI Factory, an AI ecosystem that will focus on Greek language models for health, sustainability and civilisation, powered by the DAEDALUS. This is one of seven “AI Factories” being developed across Europe. By 2023, the goal is to transform Greece into an “innovation nation” where AI enables every major digital service provided to citizens through Gov.gr.
Together, Pharos and DAEDALUS form the backbone of Greece’s AI infrastructure, while also enabling the development of AI systems tailored to the Greek language and context. Federated learning is a great fit for this – FLock.io used FL to help Sarawak train a model on the local language Sarawak Malay, which is underrepresented in traditional AI models.
Estonia is the regional leader in digitalising public services
The leading European countries in digitising public services include Malta, Estonia, Denmark, the Netherlands and Finland, featuring nearly 100 per cent online availability for government interactions and secure e-IDs. Estonia is the Baltic regional leader in digitalisation, having successfully digitised 100 per cent of its public services. Its Digital Agenda 2035 has been called the world’s most ambitious.
Estonia to build its own sovereign national AI computing infrastructure, as outlined in Estonia’s Digital Agenda 2035. Data-driven administration and automated public services will become the central operations of the state. Responding to the EU’s AI Factory programme, the Nordic AI Gigafactories project unites Estonia, Finland and Latvia (with Sweden and Denmark in negotiations to join).
The Estonian prime minister has spoken of the success of the digital state being built on trust, with digital identities, having made the country faster, simpler and more secure. Now, he is working out how to use AI in a way that makes life easier but without losing control and accountability. The next shift is towards a more agentic model of government and to become an “intelligent state”, with which come responsibilities of algorithmic transparency, data sovereignty and human accountability.
He announced on June 17, 2026 that Estonia will become the first country to create digital identities for AI agents. This will enable AI to act on behalf of people, companies or organisations within clearly defined limits and in a manner that is both verifiable and auditable.
Since Estonia is unable to build competitive alternatives to global models, authorities are deciding to instead ensure that existing models “learn” the Estonian context – an objective that federated learning is a great fit for.
The United Kingdom sits at a different crossroads
While France is cutting off Palantir in favour of a local alternative, the UK has been doing the opposite.
The controversial US software company has a growing presence in the UK and holds hundreds of millions of pounds in UK government contracts. Palantir was awarded a ÂŁ330 million contract to run the NHS Federated Data Platform, which links patient data across hospitals. In June 2026, however, the Mayor of London, Sadiq Khan, did block the Metropolitan Police from signing a deal with Palantir.
On June 3, 2026, British lawmakers called for the government to end the aforementioned NHS contract. “Palantir’s increasing presence across the public sector represents an unacceptable point of weakness” which could leave public services “at the mercy of foreign actors”, the parliamentary Commons Science, Innovation & Technology committee said in a report.
The UK sits at a contradiction: it has world-class expertise, research, supercomputing infrastructure and investment, and yet is constantly exhibiting reliance on outsourcing to foreign tech giants. It launched the Sovereign AI Fund for startups, and yet moves to award its biggest public sector contracts to Silicon Valley. Until this imbalance is amended, the nation’s massive potential to be a sovereign AI powerhouse will not be realised.
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.]
