The study (see final reference) analyzes the common characteristics to the commercial regulatory technique of regulatory sandboxes and presents its application in Spanish law through two cases: first, a general application of regulatory sandboxes in artificial intelligence systems through Royal Decree 817/2023 and, second, a sectorial application of regulatory sandboxes in the financial market through 7/2020 Act.
The study concludes with the following conclusions:
A) Common characteristics
1. The regulatory context of controlled test environments («regulatory sandboxes») is the process of digital transformation of the economy as a phenomenon of structural change that has three essential characteristics: it is inevitable, it is positive in its effects for intermediaries and consumers, and it is risky because it presents new and leveraged risks.
2. The market regulatory technique of controlled test environments («regulatory sandboxes») aims to satisfy two needs inherent to the process of digital transformation of the Economy: the convenience of promoting technological innovation and the need to protect investors in this type of projects.
3. The Spanish legal system has applied the commercial regulatory technique of controlled test environments («regulatory sandboxes») both in general areas of the economy, such as artificial intelligence, and in specific sectors, such as the financial markets and the field of competition, with the two regulatory sandboxes of the CNMC on voltage control in electricity grids.
B) Artificial Intelligence: Royal Decree 817/2023
4. The case of a general application of controlled test environments in artificial intelligence systems is regulated in Royal Decree 817/2023, of November 8.
5. Royal Decree 817/2023 is characterized by its general application and its supplementary effectiveness, because it implies a partial pre-adaptation of the future EU Regulation on artificial intelligence, by the use of the risk/opportunity method, by its immediate validity and its expiration.
6. The structure of the controlled testing environments in the field of artificial intelligence is integrated, first of all, by its subjective elements, which will be the public and private providers and users.
7. This structure is integrated, secondly, by its objective elements, which are the artificial intelligence systems and models. Artificial intelligence systems – which we qualify metaphorically as «prodigious stills» – can be classified according to two criteria: on the one hand, the criterion of the type of net information output they produce, according to which we can distinguish artificial intelligence systems and predictive systems. On the other hand, the criterion of the level of risk involved, according to which we can differentiate between high-risk AI systems and general-purpose AI systems.
8. That structure is integrated, thirdly, by its functional elements, which may be of a technical or commercial nature.
9. The functioning of the controlled environments of artificial intelligence tests is developed in two main phases: the first one is that of the fulfillment of the conditions of access of the AI providers participating in the AI DBS that goes through the stages of their selection and admission; and the second phase is that of the fulfillment of the conditions of exercise of the actions in the AI DBS by the participating providers and users that goes through the stages of the development of the programmed actions, with the required guarantees and transparency, can undergo episodes of mutation and ends with the termination of the DBS.
C) Financial markets: Act 7/2020
10. The case of a special application of controlled testing environments in financial markets is regulated in Act 7/2020, of November 13, for the digital transformation of the financial system.
11. Act 7/2020 applies the new precautionary principle which, together with the traditional principles of subsidiarity and proportionality, must inform the European regulation of innovation in technofinance.
12. Act 7/2020 should be implemented in accordance with the general EU operating principles for innovation facilitators which are: transparency, in the form of visibility of innovation facilitators to relevant market stakeholders and in the form of connectedness because innovation facilitators should provide clearly defined contact points for companies; definition: as innovation facilitators should have clearly defined and public objectives, functions and tools; traceability, because innovation facilitators should keep appropriate internal records related to their operation and continuous updating.
13. Act 7/2020 establishes a regime of regulatory sandboxes based on a subjective structure of 4 characters: the promoter of the innovation project, the participant in the project, its monitor and the competent supervisory authority.
14. The functioning of the controlled test spaces provided for in Act 7/2020 is articulated as an evolutionary process that begins with the access of the innovation project promoter to the space, continues with its development in accordance with the test protocol and ends with the exit, in case of success, through the «activity access gateway».
15. Act 7/2020 establishes the civil liability of the project promoter towards the participants, establishing its subjective elements, with the imputation of the project promoter and the exclusion of the authorities; its objective elements; and its preventive coverage through the contracting by the promoters of sufficient financial guarantees in the form of insurance, bank guarantees or bonds.
16. Act /2020 establishes a system of innovation channels or innovation hubs through transparent communication and safe harbors based on a mechanism of consultation with the competent authorities, the answers to which shall, as a general rule, be of an informative nature for the interested parties; without prejudice to the fact that, when so established by specific regulations, they shall have binding effects for the bodies and entities of the Administration in charge of the application of the corresponding rules.
Reference: Tapia Hermida, A.J., “Los entornos controlados de pruebas (regulatory sandboxes) como nueva técnica regulatoria mercantil: aplicaciones en el Derecho español”, LA LEY mercantil, n.º 109, Sección Derecho Digital. Estudios, enero 2024, La Ley, ISBN-ISSN: 2341-4537.