Noticia
Public administrations and, specifically, local entities are at a crucial moment of digital transformation. The accelerated development of artificial intelligence (AI) poses both extraordinary opportunities and complex challenges that require structured, ethical, and informed adaptation. In this context, the Spanish Federation of Municipalities and Provinces (FEMP) has launched the Practical Guide and Policies for the Use of Artificial Intelligence in Local Entities, a reference document that aspires to function as a compass for city councils, provincial councils and other local entities on their way to the responsible adoption of this technology that is advancing by leaps and bounds.
The guide is based on a key idea: AI is not just a technological issue, but also an organisational, legal, ethical and cultural one. Its implementation requires planning, governance and a strategic vision adapted to the size and digital maturity of each local authority. In this post, we will look at some of the key points in the document.
In this video you can watch the presentation session of the Guide again.
The guide is based on a key idea: AI is not only a technological issue, but also an organizational, legal, ethical and cultural one. Its implementation requires planning, governance and a strategic vision adapted to the size and digital maturity of each local entity. In this post, we will look at some relevant keys to the document.
Why an AI guide for local authorities
Local administrations have been pursuing continuous improvement of public services for years, but have often been constrained by a lack of technological resources, organizational rigidity, or data fragmentation. AI opens up an unprecedented opportunity to overcome many of these barriers, because it makes it possible to:
- Automate processes.
- Analyze large volumes of information.
- Anticipate citizen needs.
- Personalize public attention.
However, along with these opportunities come obvious risks: loss of transparency, discriminatory biases, violations of privacy or uncritical automation of decisions that affect fundamental rights. Hence the need for a guide that helps to know what can be done with AI, what should not be done and how to do it with guarantees.
48% of public administrations use Artificial Intelligence to streamline the relationship with citizens
The guide is structured around several fundamental axes that address the multiple dimensions of AI implementation at the local level:
The legal framework: the European AI Regulation as a central axis
One of the pillars of the guide is the analysis of the European Artificial Intelligence Regulation (RIA), the first comprehensive standard worldwide that regulates AI with a risk-based approach, and which fully affects local entities whether they develop, use or contract AI systems.
Specifically, the different levels of risk recognized by the RIA are:
- Unacceptable risk: includes prohibited practices such as social punctuation, subliminal manipulation or certain uses of biometrics.
- High risk: covers systems used in sensitive areas such as the management of public services, employment, education, security or administrative decision-making. These systems must meet a number of strict requirements, such as the need for human supervision or data traceability.
- Specific transparency risk: This applies to chatbots or AI-generated content and is mainly imposed on them reporting obligations (e.g. labelling content as AI-generated).
- Minimal risk: such as spam filters or AI-based video games, with no obligations, although adopting codes of conduct is recommended.
Crucially, from February 2025, local authorities must make their staff AI literate, i.e. provide training for those who operate these systems to understand their technical, legal and ethical implications. In addition, they must review whether any of their systems incur in practices prohibited by the RIA, such as subliminal manipulation or social scoring .
For local authorities, this implies the need to identify which AI systems they use, assess their level of risk and comply with the corresponding obligations.
AI, administrative procedure and data protection
The guide recalls that automating a process does not necessarily equate to using AI, but that when systems capable of inferring, recommending or deciding are incorporated, the legal impact is much greater.
The incorporation of AI in administrative procedures must respect principles such as:
- Transparency and explainability of decisions.
- Identification of the responsible body.
- Possibility of challenge.
- Effective human supervision.
In addition, the use of AI must be fully compliant with the General Data Protection Regulation (GDPR). Local authorities should be able to justify automated decisions, guarantee the rights of affected individuals and exercise extreme caution when processing sensitive personal data.
Governing Data to Govern AI
The guide is blunt on one point: AI cannot be implemented without strong data governance. AI is powered by data, and its quality, availability, and ethical management will determine the success or failure of any initiative. The document introduces the concept of "single data" and refers to unified information. In relation to this, it has been seen that many organizations discover structural deficiencies precisely when trying to implement AI, an idea that was addressed in the datos.gob.es podcast on data and AI.
For data governance in AI in the local context, the guide defines the importance of:
- Privacy by design.
- Strategic value of data.
- Institutional ethical responsibility.
- Traceability.
- Shared knowledge and quality management.
In addition, the adoption of recognized international frameworks such as DAMA-DMBOOK is recommended.
The guide also insists on the importance of the quality, availability and correct management of data to "guarantee an effective and responsible use of artificial intelligence in our local administrations". To do this, it is essential to:
• Adopt interoperability standards, such as those already existing at national and European level.
• Use APIs and secure data exchange systems, which allow information to be shared efficiently between different public bodies.
• Leverage open data sources, such as those provided by the Spanish Government's Open Data Portal or local public data platforms.
How to know the maturity status in AI
Another of the most innovative aspects of the guide is the FEMP methodology. AI, which allows local entities to self-assess their level of organizational maturity for AI deployment. This methodology distinguishes three progressive levels that should be adopted in order:
- Level 1 - Electronic Administration (EA): digitalisation of business areas through corporate solutions based on single data.
- Level 2 - Robotization of Automated Processes (RPA): automation of management processes and administrative actions.
- Level 3 - Artificial Intelligence: use of specialized analysis tools that provide information obtained automatically.
This gradual approach is essential, as it recognizes that not all entities start from the same point and that trying to skip stages can be counterproductive.
The guide stresses that AI should be seen as a support tool, not as a substitute for human judgment, especially in sensitive decisions.
Requirements for deploying AI in a local entity
The document systematically details the requirements necessary to implement AI with guarantees:
- Normative and ethical, ensuring legal compliance and respect for fundamental rights.
- Organizational, defining technical, legal, and governance roles.
- Technological, including infrastructure, integration with existing systems, scalability, and cybersecurity.
- Strategic, betting on progressive deployments, pilots and continuous evaluation.
Beyond technology, the guide underlines the importance of ethics, transparency and public trust. All this is key and points to the idea that success does not lie in advancing quickly, but in advancing well: with a solid foundation, avoiding improvisations and ensuring that AI is applied ethically, effectively and oriented to the general interest.
Likewise, the relevance of public-private collaboration and the exchange of experiences between local entities is highlighted, as a way to reduce risks, share knowledge and optimize resources.
Real cases and conclusions
The document is completed with numerous real use cases in city councils and provincial councils, which demonstrate that AI is already a tangible reality in areas such as citizen service, social management, urban planning licenses or municipal chatbots.
In conclusion, the FEMP guide is presented as an essential manual for any local entity that wants to address AI responsibly. Its main contribution is not only to explain what AI is, but to offer a practical framework to implement it in a meaningful way, always putting citizenship, fundamental rights and good governance at the centre.
