Finland and Estonia bet on AI sandboxes as Europe's small states seek influence over digital rules
Two of Europe’s most digitally advanced governments have agreed to jointly test AI systems under new EU law, positioning themselves as rule-shapers rather than rule-takers in the continent’s emerging AI governance framework.
Text by Martti Asikainen, 20.5.2026 | Photo by Adobe Stock Photos
Finland and Estonia have agreed to jointly develop regulatory sandboxes for artificial intelligence — controlled testing environments required under the EU’s Artificial Intelligence Act — following a two-day visit by a senior Estonian government delegation to Helsinki on 19–20 April 2026.
The partnership puts two small EU member states with advanced digital governments at the centre of a question that matters to the whole continent: who gets to shape the practical rules for AI deployment as the EU’s landmark legislation moves from text to implementation?
Under the AI Act, high-risk AI systems — those used in public services, employment, or critical infrastructure — must pass conformity assessments before deployment. Regulatory sandboxes let governments and companies test such systems under real conditions with reduced regulatory friction before full market entry. Agreeing on joint sandboxes means both countries can share the technical and legal groundwork rather than building duplicate national frameworks.
A longer strategic effort
The Estonian delegation was led by Secretary of State Keit Kasemets and included senior officials from across AI policy, digital governance, and economic affairs, among them the team lead of Estonia’s national AI initiative Eesti.ai. While in Helsinki, they held discussions with permanent secretaries, state secretaries, and officials from across the Finnish government.
Finland’s State Secretary Risto Artjoki described the visit as part of a longer strategic effort. “With its focus on AI and digital sovereignty, the visit is part of our countries’ long-term efforts to strengthen cooperation in a rapidly changing technology environment,” he said.
Kasemets framed the talks around both shared strengths and practical ambition. He pointed to strong data interoperability as a foundation, and to the development of what he called the “Agentic State” — the use of autonomous AI systems to deliver public services and automate government processes — and AI in education as areas where both nations are already leading. “Both our countries are uniquely positioned thanks to strong data interoperability, and there are many areas where Finland and Estonia are already frontrunners, such as the development of the Agentic State and AI in education,” he said.
He was direct about what the visit was meant to achieve. “Our goal today is simple: to identify concrete, actionable projects where we can work together quickly and pragmatically to drive economic growth and create public value,” Kasemets said.
Why these two countries
The partnership is a natural one. Estonia has built one of Europe’s most advanced digital government infrastructures over the past two decades, with data interoperability systems that have become a model for other countries. Finland, meanwhile, hosts LUMI — the most powerful supercomputer in Europe, located in Kajaani in the country’s north — and has developed a national quantum computing network, giving it significant research and infrastructure capacity.
Both are small-population EU member states, and that shared constraint is part of the logic. Neither country has the scale to act alone on the most ambitious AI projects. Together, they can move faster, share costs, and demonstrate models that larger member states may eventually adopt.
The visit also fits a broader European moment. Governments across the continent are seeking to reduce dependence on digital infrastructure and platforms controlled by non-European companies — a goal now described in Brussels as digital sovereignty.
Some questions remain
The joint press release did not specify which AI systems or policy areas the sandbox cooperation would cover first, nor did it set timelines or funding commitments beyond the general frame of the AI Act.
The reference to an Agentic State — deploying AI agents to perform government functions — signals ambition, but no specific programme was announced.
Both governments are expected to report progress on AI Act sandbox implementation by end of 2026. That deadline, more than any communiqué, will determine whether this visit produced durable policy or goodwill.
