Poland’s new AI systems act: a regulator and practical tools

On 11 June 2026 the Sejm passed Poland’s Act on artificial intelligence systems, which now goes to the Senate. The Act implements the EU AI Act (Regulation 2024/1689) – it builds the national supervisory authority and framework rather than changing the substantive obligations, which flow from the Regulation itself. For life sciences companies, especially those placing AI-enabled medical software on the Polish market, two features matter now.

Poland gains a new AI regulator, with the sector-specific authorities cooperating. The Commission for the Development and Safety of Artificial Intelligence (Komisja Rozwoju i Bezpieczeństwa Sztucznej Inteligencji – “KRiBSI”) becomes the national market-surveillance authority for AI. For life sciences com-panies this means that compliance will no longer be monitored only through sectoral regulators, for AI-related aspects. This is particularly relevant for AI-enabled medical devices, but also for pharmaceutical AI tools (such as supporting compliance in the pharmaceutical supply chain or monitoring product availa-bility). Life sciences regulators – the President of the Office for Registration of Medicinal Products, Medi-cal Devices and Biocidal Products (URPL) and the Chief Pharmaceutical Inspector (GIF) – are expressly brought into the cooperation framework with the new regulator. 

KRiBSI itself will not be operational immediately: its chair is appointed within two months of the Act enter-ing into force. The timetable leaves room to prepare, but the institutions to use it are only now being built.

Two tools are worth using before the AI Act obligations apply: 

• binding individual opinions – which may become a useful tool for borderline AI projects. The Polish implementation allows operators to request an individual opinion from KRiBSI on an exist-ing or planned factual scenario. The opinion – issued within 30 days (60 in complex cases) and for a fee of PLN 150, then binds the regulator and other state bodies for the specific facts pre-sented. The framework is intended to work similarly to the well-known tax interpretations. The mechanism could prove useful for settling an uncertain classification or compliance question for a particular product before launch. 
• regulatory sandboxes – which may be particularly relevant for digital health and medtech inno-vation. An AI regulatory sandbox is a controlled environment to test an AI system for six to twelve months, free of charge for small and medium-sized enterprises (SMEs) and with potential support in areas such as risk management documentation, training and testing datasets, tech-nical documentation, conformity assessment, including input from the data-protection authority. Both can help reduce regulatory risk on a novel AI product ahead of full compliance, but would not replace sector-specific compliance under MDR/IVDR, GDPR or clinical/regulatory rules.

Key Takeaways/ to-dos:

• Start with mapping and governance. As part of a general AI Act audit and preparation, analyse if the new tools introduced in the Polish Act may be helpful in ensuring compliance or introducing new AI-enabled solutions. 
• Consider applying to KRiBSI for a binding individual opinion to settle a classification or compli-ance question for a specific product; it binds the authorities on the facts presented, costs PLN 150 and is issued within 30 days (60 in complex cases).
• Track the launch of the Polish regulatory sandbox and the standing-up of KRiBSI: the sandbox al-lows controlled testing for six to twelve months and is free for SMEs, but the framework is still months from being operational.