The OECD.AI Policy Navigator

Our policy navigator is a living repository from more than 80 jurisdictions and organisations. Use the filters to browse initiatives and find what you are looking for.

AI Governance Framework of Madrid: Institutionalisation, Ethics and Oversight for the Public Use of Artificial Intelligence (Subnational: Madrid)

undefined
Madrid’s AI Governance establishes a corporate framework to co-ordinate, supervise and guide the responsible use of artificial intelligence across the City Council. It benefits citizens, staff and all municipal areas by supporting safety, ethics, transparency and human oversight. It was created to respond to the rapid expansion of AI and support its trustworthy use aligned with European regulation and public values.

Name in original language

Marco de Gobernanza de la IA de Madrid: Institucionalización, Ética y Supervisión para el Uso Público de la Inteligencia Artificial

Initiative overview

Madrid’s AI Governance initiative was created to address the growing need to co-ordinate AI systems and to ensure their safe, ethical and rights-based use. Before its creation, AI projects emerged in a fragmented way across departments, making it difficult to ensure transparency, oversight and alignment with European requirements. The City Council designed a comprehensive framework to guide the full lifecycle of AI systems and support a digital transformation centred on people and public value, and it set this out in the ‘Artificial Intelligence Roadmap at the Madrid City Council’.

The model has three dimensions. The procedural dimension defines norms, risk assessments, human oversight, ethical validation and data protection protocols. The structural dimension assigns roles and capabilities, with the Digital Office defining strategy, IAM as a technological enabler, and units such as Data Protection, Legal Services and Madrid Talento ensuring compliance and workforce training. The relational dimension operates through MAIA, a corporate group created in 2024 bringing together over 90 professionals to co-ordinate use cases and collaborate with universities, companies and other administrations.

The governance framework includes instruments such as the municipal algorithm registry, a standard lifecycle, ethical safeguards, legal assurance and an indicator dashboard. It aligns with the EU AI Act, the Spanish Digital Rights Charter and the OECD AI Principles, with the aim of ensuring safety, transparency, traceability, explainability and non-discrimination. It also supports continuous adaptation to the fast-changing landscape of generative AI.

Implementation is embedded within the Madrid Capital Digital strategy and materialised through more than 40 real AI projects across taxation, mobility, environment, social services, participation, tourism, urban planning, cleanliness, cybersecurity and accessibility. Examples include the Tax Laboratory, SIVPICA, PALOMA, the Mobility Big Data Platform, the Digital Twin, Decide Madrid with AI, the air quality system, the smart cleaning vehicle and VisitMadridGPT, as well as predictive models of the Madrid Transport Company (Empresa Municipal de Transporte). More information about the use cases is available on the website.

To strengthen governance, two working groups have been created: the AI Governance Working Group and the Madrid Artificial Intelligence (MAIA) Working Group. The model will evolve through a new AI Committee, expanded methodologies, a reinforced algorithm registry, periodic audits, stronger alignment with the AI Act and increased citizen engagement. It is designed as a permanent, scalable and replicable governance mechanism.

Other relevant details

Results, outcomes and impacts: AI governance has enabled co-ordination of more than 40 use cases across key areas, supporting ethical, technical and legal consistency. MAIA, with more than 90 professionals, and a transversal model embedded in Madrid Capital Digital, are already operational. Reported use-case results include: 6,622 asbestos roofs detected (94.2% accuracy), 5,000 calls and 646 follow-ups in PALOMA, 2.7 million air quality data downloads, 5.7 million interactions in Línea Madrid, a 50% reduction in cleaning supervision time and 92% accuracy in waste detection. Challenges and lessons learned: Challenges included heterogeneous data, an initial lack of common criteria, technical complexity, ethical risks (bias, privacy, explainability) and workforce capacity gaps. Co-ordinating departments with uneven digital maturity and aligning with the EU AI Act also required effort. The City Council addressed these through a structured governance model, standard processes, MAIA, ethical guidelines, legal assurance, human oversight, an algorithm registry and extensive training. Lessons include the value of transversal governance, early expert involvement, continuous evaluation and transparent documentation.