Spain legislates for first EU AI Act regulatory sandbox

Technology law experts Paloma Bru and Lidia Vidal of Pinsent Masons in Madrid said that Spanish Royal Decree 817/2023 (31-page / 391KB PDF) issued on 8 November highlights the Spanish government’s intention to play a leading role in promoting innovation in AI.

Paloma Bru said: “According to the proposed new EU AI Act, member states are encouraged to launch AI regulatory sandboxes to create a controlled environment for testing AI systems against the requirements of the AI Act under the direct supervision of competent authorities. Spain is leading these efforts, having approved the first EU sandbox aimed at testing the AI regulatory requirements.”

The new sandbox provided for in the Royal Decree envisages cooperation between providers of AI systems and their users and, among other things, is designed to enable private and public sector organisations to test their AI systems against requirements of the AI Act before the AI Act begins to take effect. Participation in the sandbox is dependent on meeting eligibility criteria.

The operation of Spain’s sandbox, which is backed by the European Commission, is expected to help inform regulatory compliance and the development of new technical guidance.

“As the ecosystem prepares for the approval of the EU AI Act, it is expected that guidelines and best practices that prepare and raise awareness among companies will be produced, especially among SMEs and start-ups, to facilitate their implementation,” Lidia Vidal said.

“The initiative is expected to lead to the development of a report with best practices, lessons learned, and technical application guidelines based on testing and practical experimentation. These deliverables can be then used by the European Commission for the development of EU guidelines, facilitating compliance with the future AI regulation by companies, especially SMEs,” she said.

The Royal Decree also establishes the creation of an AI advisory committee made up of independent technical experts.

Paloma Bru said the new measures need to be seen in the wider context of Spain’s digital transformation and artificial intelligence strategies.

In August, Spanish Royal Decree 729/2023 was separately issued to provide for the establishment of the new Spanish Agency for the Supervision of Artificial Intelligence (AESIA). It is the first AI regulatory body of this kind appointed in the EU in compliance with the proposed EU AI Act and is expected to begin operating in early December.

Lidia Vidal said: “The creation of AESIA reflects the compliance with the obligation of the AI Act for EU member states to designate a regulator responsible for enforcing the AI Act and acting as a single point of contact for the European Commission. Among other objectives, it will promote real test environments for AI systems, to reinforce user protection and avoid discriminatory biases. It will also have the capacity to impose fines – potentially to 6% of a company’s total worldwide annual turnover for the preceding financial year, under some proposals for the AI Act.”

“AESIA will not replace the role currently played by the Spanish data protection authority (AEPD) with respect to AI, which will retain responsibility for overseeing compliance with Spanish data protection laws in the context of AI-based processing of personal data. However, the AEPD and the AESIA will have to collaborate and work together given the intersection between the proposed new AI Act and the GDPR,” Paloma Bru said.

Spain currently holds the presidency of the Council of Ministers, which, along with the European Parliament, is responsible for scrutinising and adopting proposed new EU legislation. Paloma Bru said the Spanish government is pushing for the Council and Parliament to adopt the proposed new EU AI Act before its presidency comes to an end at the end of this year.

“Spain has been playing an active role and taking significant steps to position itself as a leading country in AI, demonstrating its commitment to encourage and ensure its proper regulation,” Lidia Vidal said. “However, it may be a challenge to reach a deal on the AI Act before the end of this year due to reported roadblocks in the negotiations.”