EU AI Act – Latest Regulatory Developments in Finland

Author: Maria Aholainen

What has happened so far?

In Finland, the EU AI Act will be transposed into Finnish law by way of two government proposals:

1. On 8 May 2025, the Government submitted a Government Proposal 46/2025 to the Parliament. The Government proposes a new Act on the Supervision of Certain AI Systems, with a focus on supervisory authorities and penalties in Finland.
2. On 25 April 2025, the Ministry of Economic Affairs and Employment published a separate Draft Government Proposal on AI Sandboxes in Finland.

While the EU AI Act is directly applicable in all EU Member States, including Finland, each Member State must supplement the EU AI Act by designating supervisory authorities with powers to impose administrative fines and other penalties. In addition, the EU AI Act provides EU Member States with limited leeway, i.e. flexibility in tailoring the national rules.

Who is responsible for supervision in Finland?

Finland has chosen a decentralised supervision. Instead of one single authority, several market surveillance authorities are tasked with supervision of AI systems — the authorities who are already today responsible for the supervision in the field of product safety, road traffic, digital infrastructure, medical devices, and financial services, etc. The Finnish Transport and Communications Agency (“Traficom”) will act as the single point of contact under the EU AI Act and will be responsible for the coordination between national authorities.

Compared to the draft government proposal (published in October 2024), there are no major changes. The decentralised model was already proposed in the draft version, and in the final proposal, the competences between several supervisory authorities have been redefined based on their comments. In addition, attention has been paid to cooperation between the supervisory authorities to ensure coherent interpretation and enforcement. Based on the feedback received during the public consultation, the guidance and coordination by and between the supervisory authorities were widely seen as prerequisites for successful national implementation.

How about penalties?

A new Sanctions Board will be established with the powers to impose administrative fines higher than EUR 100,000 whereas fines under EUR 100,000 are imposed by individual supervisory authorities.

The sanction threshold in the draft proposal was EUR 300,000, and it has now been reduced to EUR 100,000. Further, the minimum fine has been omitted from the final proposal.

How about national leeway?

National leeway is hardly used. However, we note that Finland has (a lot of!) national legislation that may restrict the use of AI in Finland, especially in the public sector. The Chancellor of Justice has raised this issue in the Preliminary Review, and a clarification has been added to the government proposal stating that the public sector cannot use AI systems for automated decision-making due to national legislation. This would require changes to national legislation in line with Article 22 of GDPR.

AI sandboxes – what are AI sandboxes and how are they established in Finland?

The AI Act enables the establishment of AI regulatory sandboxes to provide a controlled environment in which to test and validate AI systems before they are placed on the market. The purpose is primarily to support small and medium enterprises. Each Member State (alone or together with other Member States) must establish at least one AI sandbox under the EU AI Act.

Based on the draft proposal, there will be one AI sandbox (test environment under the EU AI Act) in Finland and Traficom is the supervisory authority for the sandbox and responsible for cooperation and coordination between supervisory authorities. In addition to the AI sandbox, the draft proposal covers high-risk AI systems in the area of critical infrastructure that should be registered at national level.

What are the next steps in Finland?

The Government Proposal on the Act on the Supervision of Certain AI Systems is in the Parliament and is expected to enter into force on 2 August 2025.

The draft proposal on AI sandboxes was sent out for public consultation, and the deadline for comments is 2 June 2025. The rules on AI sandboxes are expected to enter into force on 1 February 2026.

Contact us

Maria Aholainen