Poland seeks delay of EU AI Act penalties for high-risk systems
Warsaw proposed suspending sanctions under Article 99 while keeping oversight powers, expanding the use of AI sandboxes, and providing clearer compliance guidance.
Poland has proposed a delay to the EU AI Act’s penalty regime for high-risk artificial intelligence systems, raising questions about how the landmark regulation will be applied in its early stages. In a non-paper circulated ahead of next week’s Telecom Working Party, Warsaw called for sanctions under Article 99 of the Act to be suspended for six to twelve months, while keeping national enforcement powers in place. The move would allow authorities to monitor and oversee compliance, but without imposing fines during the initial rollout.
The proposal argues that companies, particularly small and medium-sized enterprises, face significant challenges in preparing for the new rules. Poland points to the absence of finalised technical standards and guidance as a barrier to compliance, and warns that smaller businesses could otherwise face disproportionate costs or consider relocating to jurisdictions with lighter regulation. To address this, Warsaw has urged the European Commission and member states to provide clearer interpretative guidance and to recognise alternative solutions where standards are not yet available.
Another element of the proposal is the expanded use of AI sandboxes—controlled environments where companies can develop and test systems under regulatory supervision. Poland suggests that sandboxes could ease the transition by allowing firms to adapt to the rules in practice before being exposed to full liability. The debate over the timing of penalties and the possible introduction of a “stop the clock” mechanism will be a central issue at the upcoming Telecom Working Party meeting, where EU countries are expected to weigh the need for regulatory certainty against concerns over innovation, competitiveness, and fairness for smaller actors.
