The EU AI Act enters a new phase. More obligations await Poland and Europe
In August 2025, further provisions of the EU’s Artificial IntelligenceAct (AI Act) came into force, introducing significant new requirements for technology providers and member state administrations. The regulations focus on general purpose AI (GPAI) models and the creation of surveillance structures. Poland, like other EU countries, is working to adapt its national law, although work on a key law is still ongoing.
A new phase in the implementation of the AI Act, a landmark regulation for the technology sector, focuses on ensuring the transparency and security of the most advanced AI systems. Providers of general-purpose models, such as those underpinning popular chatbots or image generators, have been required to assess their impact on fundamental rights in detail. They must also publish concise summaries of the data used to train the algorithms and provide oversight of their further use.
This is the regulator’s response to growing concerns about the potential risks associated with the non-transparent operation of powerful language and generative models. The European Commission has also published a specific code of conduct to help companies implement these complex requirements in practice.
Another pillar of the August amendments is the establishment of a formal governance and oversight framework for the AI market. The legislation establishes EU administrative structures which, in cooperation with national authorities, will monitor compliance with the law. A system of financial penalties for violations has also been introduced to ensure real enforcement of the regulations.
Meanwhile, in Poland, government work is underway on a draft Artificial Intelligence Systems Act (U71) to implement the EU law. The draft, which has received a positive opinion from the Committee for European Affairs, envisages the establishment of a new supervisory body, the Commission for the Development and Security of Artificial Intelligence. The Ministry of Digitalisation has a key role in coordinating the implementation.
An important element of the Polish strategy is to be the promotion of innovation through the creation of so-called regulatory sandboxes. These controlled test environments will allow companies and startups to experiment with new AI technologies without the risk of violating complex regulations, which is supposed to be a compromise between security and development. However, the lack of finalisation of national regulations remains a challenge for the Polish AI sector, which is waiting for clear implementation guidelines.
