Artificial Intelligence - Sweden

The EU AI Act (Regulation 2024/1689) is directly applicable in Sweden. Prohibitions on unacceptable-risk AI (Article 5) applied from 2 February 2025; GPAI model rules from 2 August 2025; high-risk AI system obligations fully applicable from 2 August 2026 (extended further under the 2026 Digital Omnibus agreement).

Sweden's Government published its first comprehensive national AI Strategy in February 2026, aiming to place Sweden among the world's top-10 AI nations by 2030. The strategy sets priorities for AI in public administration, competitiveness, responsible development, and secure infrastructure, accompanied by a detailed action plan.

Sweden's Government-appointed AI Commission, led by Carl-Henric Svanberg, submitted 75 policy proposals in December 2024 covering public-sector AI uptake, innovation facilitation, and safe/ethical AI development. The majority of measures were incorporated into government policy in 2025 and fed directly into the 2026 AI Strategy.

The official government inquiry SOU 2025:101 (Anpassningar till AI-förordningen) proposes designating PTS as the lead national competent authority and market surveillance authority for the EU AI Act, with IMY holding primary responsibility for prohibited-practices enforcement (Article 5) and Finansinspektionen (FI) jointly covering financial-sector AI. Nine additional sector authorities are proposed for product-safety domains. PTS would also host the national AI regulatory sandbox.

On 21 January 2025, DIGG (Agency for Digital Government) and IMY (Data Protection Authority) jointly published 18 national guidelines for generative AI use in Swedish public administration, covering management responsibility, information security, copyright, GDPR compliance, ethics, labour law, and procurement. These are non-binding but government-mandated guidance.

A political agreement between the EU Council and Parliament reached on 7 May 2026 streamlines and simplifies the EU AI Act's high-risk AI obligations. Under the amended timeline, stand-alone high-risk AI systems must comply by 2 December 2027, and product-embedded high-risk AI systems by 2 August 2028 — extending the deadlines that directly affect Swedish operators.