Spain legislates the EU's first AI Act Regulatory Sandbox

Spain legislates the EU's first AI Act Regulatory Sandbox

Spain has launched the first AI Act Regulatory Sandbox in the European Union, providing a controlled environment for testing AI systems. The legislation on the sandbox, announced on November 8, 2023, allows participants to evaluate their AI systems' compliance with the proposed EU AI Act and to undergo self-assessment and post-marketing evaluation.

For what and who can the sandbox be used?

The AI Act Regulatory Sandbox is designed for high-risk AI systems, general purpose systems (like ChatGPT), and foundation models. Projects must focus on innovation, compliance testing, and preparation for market introduction, while ensuring data protection if personal data is processed. Certain AI systems, such as those used for military or national security purposes, or those exploiting specific group vulnerabilities, are excluded from participating.

Participants of the Spanish project must be permanently established in Spain or be part of an entity group with a representative in Spain. This includes both private entities (ranging from local startups to international companies with a Spanish branch) and government agencies.

(Dis)advantages and alternatives

The main advantage for AI developers is receiving feedback on compliance from regulators, potentially leading to a sort of 'seal of approval' for their systems. However, participation requires significant time and resource investment, and there's a risk of a negative exit report.

An alternative to the sandbox is 'real world testing', which allows for testing AI systems under actual conditions with real users and data. It requires an approved test plan and adherence to various conditions, including ethical assessment and informed consent of participants.

Paving the way

The Spanish regulation could serve as a model for other EU countries, including the Netherlands, especially for developers of high-risk AI systems. It's less suitable for small businesses with less resources and those working on low-risk AI applications.Healthcare, financial, and government institutions could particularly benefit from a Dutch AI sandbox, emphasizing privacy, security, and ethics.

In summary, Spain's AI Act Regulatory Sandbox initiative provides an unique opportunity for AI developers to test their systems for future EU AI regulations and gain experience in conformity assessments and post-market surveillance. While it involves investments and risks, promoting responsible technological advancement as well.