Spain - Asturias - AI Regulation (98/2025)

Overview

The Decree 98/2025, officially titled 'Decreto 98/2025, de 22 de julio, por el que se regula el uso de la Inteligencia Artificial en la Administración del Principado de Asturias y su sector público,' represents a landmark legislative effort by the Principality of Asturias, Spain. Enacted on July 22, 2025, and subsequently published in the Boletín Oficial del Principado de Asturias (BOPA) on July 31, 2025, this decree entered into force three months later, on October 31, 2025. Its primary objective is to establish a comprehensive and pioneering regulatory framework for the responsible, ethical, and secure deployment of Artificial Intelligence (AI) systems within the regional public administration and its broader public sector. This initiative positions Asturias at the forefront of AI governance within Spain and Europe, demonstrating a proactive approach to managing the transformative impact of AI on public services.

The decree is designed with a dual purpose: first, to guarantee that all AI applications utilized by the Asturian administration adhere strictly to ethical principles, ensure security, and respect fundamental rights; and second, to foster the responsible and controlled development of AI innovation. It seeks to leverage AI to improve the quality, personalization, and proactivity of public services, particularly in critical sectors such as healthcare, social services, and general administrative management. By doing so, the Principality aims to enhance the relationship between citizens and the administration, streamline bureaucratic processes, and optimize public resource allocation, all while maintaining robust public control and accountability over AI systems.

Definitions

While the full text of the Decree 98/2025 is required for precise legal definitions, the overarching themes and specific mentions in official communications and legal analyses provide insight into the key terms addressed. 'Inteligencia Artificial (IA)' refers broadly to systems designed to operate with a degree of autonomy, capable of processing data, learning, reasoning, and making decisions or predictions to achieve specific objectives. The decree specifically targets the 'Administración del Principado de Asturias y su sector público,' defining the scope of entities subject to its regulations, encompassing various regional government bodies and public enterprises.

A central concept introduced is the 'sandbox de inteligencia artificial' or 'entorno de pruebas controlado,' which designates a secure and regulated environment where new AI systems can be tested and evaluated before their definitive implementation in public services. This mechanism is crucial for identifying potential errors, biases ('sesgos'), and risks, ensuring data security, and verifying alignment with ethical and legal principles. Other important terms include 'gobernanza' (governance), referring to the institutional structures and processes for oversight; 'supervisión humana,' emphasizing the necessity of human intervention and control over automated decisions; 'transparencia' and 'explicabilidad,' requiring AI systems to be understandable and their outputs justifiable; and 'ciberseguridad,' highlighting the need to protect AI systems and the data they process from cyber threats.

Governance and Institutional Framework

The Decree 98/2025 establishes a clear governance structure for the oversight and control of AI systems within the Asturian public administration. The primary responsibility for authorizing, supervising, and controlling AI systems falls under the 'Consejería competente en materia de IA,' which is the regional Ministry with jurisdiction over Artificial Intelligence. This central authority ensures a coordinated and consistent approach to AI deployment across various public entities. Complementing this, the 'Dirección General competente en materia de IA' (Directorate General for AI) is tasked with the evaluation of compliance requirements for AI systems, as well as the ongoing supervision and control of their results.

Further institutional support is provided by the 'CEDISI' (Centro de Servicios Compartidos del Principado de Asturias), which assumes new functions related to AI. CEDISI is expected to play a strategic role in areas such as training, evaluation, and dissemination of best practices concerning AI. This collaborative governance model aims to ensure that expertise is centralized while implementation is distributed, fostering a responsible and innovative ecosystem. The framework also emphasizes that these regional competencies operate without prejudice to the broader supervision and market surveillance responsibilities held by various national authorities, in accordance with the provisions of the European AI Regulation.

Key Focus Areas

The Decree 98/2025 places significant emphasis on several key areas to ensure the responsible and beneficial integration of AI into public services. A cornerstone is the protection of 'derechos fundamentales' (fundamental rights), ensuring that AI systems do not infringe upon individual liberties, promote discrimination, or perpetuate biases. This is achieved through strict requirements for data quality, mandating that data used for AI training and operation must be adequate, free from errors, and devoid of biases, while complying with the General Data Protection Regulation (GDPR).

Another critical focus is 'transparencia' and 'explicabilidad,' requiring AI systems to be understandable, their decision-making processes traceable, and their outcomes justifiable to human operators and citizens. This principle ensures accountability and builds trust in AI-driven public services. The decree also mandates 'supervisión humana,' ensuring that human oversight is maintained, and automated decisions do not override human autonomy or judgment, particularly in sensitive contexts. Furthermore, 'ciberseguridad' is a paramount concern, with provisions to guarantee the security and resilience of AI systems against attacks and vulnerabilities. The decree explicitly prohibits practices deemed unacceptable by the European AI Act, such as harmful cognitive-behavioral manipulation or social scoring by authorities, reinforcing a robust ethical stance.

Implementation Framework

The implementation framework outlined in Decree 98/2025 is designed to facilitate the secure and responsible deployment of AI systems within the Asturian public administration. A crucial component is the requirement for 'autorización previa' (prior authorization) before any AI system can be deployed. This authorization process involves a thorough assessment by the competent Consejería, ensuring that all regulatory and ethical requirements are met before an AI system becomes operational. This centralized approval mechanism provides a vital gatekeeping function, preventing the hasty or unvetted introduction of AI technologies.

A distinctive feature of the implementation framework is the establishment of 'entornos controlados de prueba' or 'sandboxes.' These regulatory sandboxes provide a safe and controlled environment for testing AI systems, allowing for the verification, monitoring, and evaluation of their performance, efficacy, and adherence to ethical guidelines before full-scale deployment. This phased approach enables the detection and correction of errors or biases, minimizing potential risks to citizens and public services. The decree also actively promotes 'fomento de la innovación' by encouraging the participation of small and medium-sized enterprises (SMEs) and startups in these testing environments and in public procurement processes, offering favorable conditions to integrate cutting-edge AI solutions into the public sector.

Monitoring and Evaluation

The Decree 98/2025 recognizes the dynamic nature of AI technologies and the necessity for continuous oversight post-deployment. Consequently, it mandates a robust framework for 'monitorización técnica' and ongoing evaluation of AI systems once they are in operation within the public administration. This continuous monitoring aims to track the performance, accuracy, and adherence to ethical principles of AI systems over their lifecycle, ensuring that they continue to meet regulatory standards and deliver intended benefits without unforeseen negative consequences.

A critical aspect of this framework is the provision for 'evaluación a posteriori del impacto y rendimiento' (post-implementation evaluation of impact and performance). This involves periodic assessments to determine the actual effects of AI systems on public services, citizens, and administrative efficiency, allowing for adjustments and improvements as needed. Furthermore, the decree includes a 'capacidad de retirada' (capacity for withdrawal), meaning that AI systems can be suspended temporarily or permanently if they present errors, biases, or artificial 'hallucinations' that compromise their reliability, safety, or ethical compliance. This mechanism ensures that the administration can swiftly respond to and mitigate any identified risks, upholding public trust and safeguarding fundamental rights.

Penalties, Liability, and Appeals

While the Decree 98/2025 primarily focuses on establishing a framework for responsible AI deployment and oversight within the Asturian public administration, the specific details regarding penalties, liability, and appeal mechanisms within the decree itself are not extensively detailed in the provided search snippets. However, the decree's strong alignment with the European Union's AI Act (Regulation (EU) 2024/1689) implies that the broader European and national sanctioning frameworks would apply. The EU AI Act includes provisions for significant penalties for non-compliance, with prohibitions becoming effective in February 2025 and governance rules applicable from August 2025.

The decree does explicitly mention that AI systems can be withdrawn or suspended if they present risks, errors, or biases, indicating a direct consequence for non-compliant or malfunctioning systems within the public sector. For instance, if an AI system fails to guarantee cybersecurity, exhibits uncorrected biases, or poses risks, its deployment can be halted. It is reasonable to infer that severe breaches of the decree's principles, particularly those related to fundamental rights or data protection, would fall under existing Spanish administrative law and potentially the sanctioning powers established by the EU AI Act and any subsequent national implementing legislation. A Spanish national draft law for the good use and governance of AI, approved in March 2025, for example, proposes fines of up to €35 million for failing to label AI-generated content, indicating the direction of national enforcement. The Asturian decree sets the regional operational standards that would likely inform the application of such broader penalties.

Relationship to Other Instruments

The Decree 98/2025 is intricately linked to and largely inspired by the European Union's comprehensive legal framework for Artificial Intelligence, specifically Regulation (EU) 2024/1689, commonly known as the EU AI Act. The Asturian decree explicitly states its alignment with this EU regulation, aiming to adapt the necessary normative measures at a regional level to ensure that the EU AI Act effectively deploys its effects and contributes to its overarching objectives. This relationship positions the Asturian decree not as an isolated piece of legislation, but as a complementary instrument designed to operationalize and localize the broader European principles and requirements for AI governance within the specific context of the Principality's public administration.

The EU AI Act, which entered into force on August 1, 2024, with various provisions becoming applicable in phases from February 2025 to August 2027, establishes harmonized rules for AI across the Union, employing a risk-based approach. The Asturian decree mirrors this approach by focusing on ethical use, transparency, human oversight, and risk management, particularly for AI systems used in critical public services. By anticipating the full application of the EU AI Act, the Principality of Asturias demonstrates its commitment to a unified European approach to AI regulation, while also tailoring specific implementation mechanisms, such as the AI sandbox, to its regional administrative needs and innovative aspirations.

International Alignment

The international alignment of the Asturias AI Decree is primarily and strongly with the European Union's regulatory framework for Artificial Intelligence. As a regional decree within Spain, a Member State of the European Union, its foundational principles and specific provisions are designed to be consistent with, and indeed to facilitate the implementation of, the EU AI Act (Regulation (EU) 2024/1689). This alignment ensures that the Asturian public administration operates within a harmonized European standard for AI governance, promoting trust, safety, and ethical considerations across borders within the EU.

The EU AI Act itself is considered the world's first comprehensive law on AI, setting a global precedent for responsible AI development and deployment. By explicitly referencing and adapting to this regulation, the Asturias AI Decree indirectly aligns with the broader international discourse on AI ethics and governance, which often draws inspiration from the EU's risk-based approach. This includes shared principles such as the protection of fundamental rights, transparency, human oversight, data quality, and accountability, which are increasingly recognized as essential for trustworthy AI globally. The decree's emphasis on fostering innovation while mitigating risks also resonates with international efforts to balance technological advancement with societal well-being.

Implementation Timeline

Sources and References