How Can AI regulatory sandboxes realise their full potential under the AI Act?
With the AI Act, Europe aims to strike a careful balance: protecting fundamental rights while promoting responsible innovation — especially when it comes to high-risk AI systems. AI regulatory sandboxes, introduced under Article 57 of the Act, can be a key tool to achieve this ambition.
AI Regulatory Sandboxes offer more than legal flexibility. They create protected spaces where innovators, regulators and civil society can experiment, learn and build trust — together.
Our new policy paper explores how sandboxes can be designed and implemented to truly support the goals of the AI Act.
What you’ll find in the paper
- A clear overview of the role and purpose of AI regulatory sandboxes within Europe’s AI governance framework
- Seven key conditions for designing and running effective sandboxes
- Actionable recommendations for policy makers at both EU and national level
- Guidance on institutional anchoring, collaboration models and knowledge-sharing mechanisms
Why it matters now
By August 2026, every EU member state is required to have a fully operational AI regulatory sandbox in place. Preparing early is crucial — not just for meeting this deadline, but for laying the groundwork for a robust, trustworthy and innovation-friendly AI landscape across Europe.
This paper offers timely insights — supporting the European Commission’s work on the upcoming implementing act (Article 58), while also guiding national-level implementation.
It is both a strategic impulse and an invitation: to collaborate, to learn, and to shape.
Download the full policy paper
An impetus for the strategic use of real-world laboratories for a sustainable AI ecosystem in Europe.
Lajla Fetic
Head of AI Public Interest & Policy
Demian Niemeyer
Dr. Till Klein
Head of Trustworthy AI