Ciencia y tecnología

Data is a fundamental resource for improving our quality of life because it enables better decision-making processes to create personalised products and services, both in the public and private sectors. In contexts such as health, mobility, energy or education, the use of data facilitates more efficient solutions adapted to people's real needs. However, in working with data, privacy plays a key role. In this post, we will look at how data spaces, the federated computing paradigm and federated learning, one of its most powerful applications, provide a balanced solution for harnessing the potential of data without compromising privacy. In addition, we will highlight how federated learning can also be used with open data to enhance its reuse in a collaborative, incremental and efficient way.

Privacy, a key issue in data management

As mentioned above, the intensive use of data requires increasing attention to privacy. For example, in eHealth, secondary misuse of electronic health record data could violate patients' fundamental rights. One effective way to preserve privacy is through data ecosystems that prioritise data sovereignty, such as data spaces. A dataspace is a federated data management system that allows data to be exchanged reliably between providers and consumers. In addition, the data space ensures the interoperability of data to create products and services that create value. In a data space, each provider maintains its own governance rules, retaining control over its data (i.e. sovereignty over its data), while enabling its re-use by consumers. This implies that each provider should be able to decide what data it shares, with whom and under what conditions, ensuring compliance with its interests and legal obligations.

Federated computing and data spaces

Data spaces represent an evolution in data management, related to a paradigm called federated computing, where data is reused without the need for data flow from data providers to consumers. In federated computing, providers transform their data into privacy-preserving intermediate results so that they can be sent to data consumers. In addition, this enables other Data Privacy-Enhancing Technologies(Privacy-Enhancing Technologies)to be applied. Federated computing aligns perfectly with reference architectures such as Gaia-X and its Trust Framework, which sets out the principles and requirements to ensure secure, transparent and rule-compliant data exchange between data providers and data consumers.

Federated learning

One of the most powerful applications of federated computing is federated machine learning ( federated learning), an artificial intelligence technique that allows models to be trained without centralising data. That is, instead of sending the data to a central server for processing, what is sent are the models trained locally by each participant.

These models are then combined centrally to create a global model. As an example, imagine a consortium of hospitals that wants to develop a predictive model to detect a rare disease. Every hospital holds sensitive patient data, and open sharing of this data is not feasible due to privacy concerns (including other legal or ethical issues). With federated learning, each hospital trains the model locally with its own data, and only shares the model parameters (training results) centrally. Thus, the final model leverages the diversity of data from all hospitals without compromising the individual privacy and data governance rules of each hospital.

Training in federated learning usually follows an iterative cycle:

1. A central server starts a base model and sends it to each of the participating distributed nodes.
2. Each node trains the model locally with its data.
3. Nodes return only the parameters of the updated model, not the data (i.e. data shuttling is avoided).
4. The central server aggregates parameter updates, training results at each node and updates the global model.
5. The cycle is repeated until a sufficiently accurate model is achieved.

Figure 1. Visual representing the federated learning training process. Own elaboration

This approach is compatible with various machine learning algorithms, including deep neural networks, regression models, classifiers, etc.

Benefits and challenges of federated learning

Federated learning offers multiple benefits by avoiding data shuffling. Below are the most notable examples:

1. Privacy and compliance: by remaining at source, data exposure risks are significantly reduced and compliance with regulations such as the General Data Protection Regulation (GDPR) is facilitated.
2. Data sovereignty: Each entity retains full control over its data, which avoids competitive conflicts.
3. Efficiency: avoids the cost and complexity of exchanging large volumes of data, speeding up processing and development times.
4. Trust: facilitates frictionless collaboration between organisations.

There are several use cases in which federated learning is necessary, for example:

• Health: Hospitals and research centres can collaborate on predictive models without sharing patient data.
• Finance: banks and insurers can build fraud detection or risk-sharing analysis models, while respecting the confidentiality of their customers.
• Smart tourism: tourist destinations can analyse visitor flows or consumption patterns without the need to unify the databases of their stakeholders (both public and private).
• Industry: Companies in the same industry can train models for predictive maintenance or operational efficiency without revealing competitive data.

While its benefits are clear in a variety of use cases, federated learning also presents technical and organisational challenges:

• Data heterogeneity: Local data may have different formats or structures, making training difficult. In addition, the layout of this data may change over time, which presents an added difficulty.
• Unbalanced data: Some nodes may have more or higher quality data than others, which may skew the overall model.
• Local computational costs: Each node needs sufficient resources to train the model locally.
• Synchronisation: the training cycle requires good coordination between nodes to avoid latency or errors.

Beyond federated learning

Although the most prominent application of federated computing is federated learning, many additional applications in data management are emerging, such as federated data analytics (federated analytics). Federated data analysis allows statistical and descriptive analyses to be performed on distributed data without the need to move the data to the consumers; instead, each provider performs the required statistical calculations locally and only shares the aggregated results with the consumer according to their requirements and permissions. The following table shows the differences between federated learning and federated data analysis.

Figure 1. Comparative table. Source: own elaboration

Federated learning and open data: a symbiosis to be explored

In principle, open data resolves privacy issues prior to publication, so one would think that federated learning techniques would not be necessary. Nothing could be further from the truth. The use of federated learning techniques can bring significant advantages in the management and exploitation of open data. In fact, the first aspect to highlight is that open data portals such as datos.gob.es or data.europa.eu are federated environments. Therefore, in these portals, the application of federated learning on large datasets would allow models to be trained directly at source, avoiding transfer and processing costs. On the other hand, federated learning would facilitate the combination of open data with other sensitive data without compromising the privacy of the latter. Finally, the nature of a wide variety of open data types is very dynamic (such as traffic data), so federated learning would enable incremental training, automatically considering new updates to open datasets as they are published, without the need to restart costly training processes.

Federated learning, the basis for privacy-friendly AI

Federated machine learning represents a necessary evolution in the way we develop artificial intelligence services, especially in contexts where data is sensitive or distributed across multiple providers. Its natural alignment with the concept of the data space makes it a key technology to drive innovation based on data sharing, taking into account privacy and maintaining data sovereignty.

As regulation (such as the European Health Data Space Regulation) and data space infrastructures evolve, federated learning, and other types of federated computing, will play an increasingly important role in data sharing, maximising the value of data, but without compromising privacy. Finally, it is worth noting that, far from being unnecessary, federated learning can become a strategic ally to improve the efficiency, governance and impact of open data ecosystems.

Jose Norberto Mazón, Professor of Computer Languages and Systems at the University of Alicante. The contents and views reflected in this publication are the sole responsibility of the author.

In the current landscape of data analysis and artificial intelligence, the automatic generation of comprehensive and coherent reports represents a significant challenge. While traditional tools allow for data visualization or generating isolated statistics, there is a need for systems that can investigate a topic in depth, gather information from diverse sources, and synthesize findings into a structured and coherent report.

In this practical exercise, we will explore the development of a report generation agent based on LangGraph and artificial intelligence. Unlike traditional approaches based on templates or predefined statistical analysis, our solution leverages the latest advances in language models to:

1. Create virtual teams of analysts specialized in different aspects of a topic.
2. Conduct simulated interviews to gather detailed information.
3. Synthesize the findings into a coherent and well-structured report.

Access the data laboratory repository on Github.

Run the data preprocessing code on Google Colab.

As shown in Figure 1, the complete agent flow follows a logical sequence that goes from the initial generation of questions to the final drafting of the report.

Figure 1. Agent flow diagram.

Application Architecture

The core of the application is based on a modular design implemented as an interconnected state graph, where each module represents a specific functionality in the report generation process. This structure allows for a flexible workflow, recursive when necessary, and with capacity for human intervention at strategic points.

Main Components

The system consists of three fundamental modules that work together:

1. Virtual Analysts Generator

This component creates a diverse team of virtual analysts specialized in different aspects of the topic to be investigated. The flow includes:

• Initial creation of profiles based on the research topic.
• Human feedback point that allows reviewing and refining the generated profiles.
• Optional regeneration of analysts incorporating suggestions.

This approach ensures that the final report includes diverse and complementary perspectives, enriching the analysis.

2. Interview System

Once the analysts are generated, each one participates in a simulated interview process that includes:

• Generation of relevant questions based on the analyst's profile.
• Information search in sources via Tavily Search and Wikipedia.
• Generation of informative responses combining the obtained information.
• Automatic decision on whether to continue or end the interview based on the information gathered.
• Storage of the transcript for subsequent processing.

The interview system represents the heart of the agent, where the information that will nourish the final report is obtained. As shown in Figure 2, this process can be monitored in real time through LangSmith, an open observability tool that allows tracking each step of the flow.

Figure 2. System monitoring via LangGraph. Concrete example of an analyst-interviewer interaction.

3. Report Generator

Finally, the system processes the interviews to create a coherent report through:

• Writing individual sections based on each interview.
• Creating an introduction that presents the topic and structure of the report.
• Organizing the main content that integrates all sections.
• Generating a conclusion that synthesizes the main findings.
• Consolidating all sources used.

The Figure 3 shows an example of the report resulting from the complete process, demonstrating the quality and structure of the final document generated automatically.

Figure 3. View of the report resulting from the automatic generation process to the prompt "Open data in Spain".

What can you learn?

This practical exercise allows you to learn:

Integration of advanced AI in information processing systems:

• How to communicate effectively with language models.
• Techniques to structure prompts that generate coherent and useful responses.
• Strategies to simulate virtual teams of experts.

Development with LangGraph:

• Creation of state graphs to model complex flows.
• Implementation of conditional decision points.
• Design of systems with human intervention at strategic points.

Parallel processing with LLMs:

• Parallelization techniques for tasks with language models.
• Coordination of multiple independent subprocesses.
• Methods for consolidating scattered information.

Good design practices:

• Modular structuring of complex systems.
• Error handling and retries.
• Tracking and debugging workflows through LangSmith.

Conclusions and future

This exercise demonstrates the extraordinary potential of artificial intelligence as a bridge between data and end users. Through the practical case developed, we can observe how the combination of advanced language models with flexible architectures based on graphs opens new possibilities for automatic report generation.

The ability to simulate virtual expert teams, perform parallel research and synthesize findings into coherent documents, represents a significant step towards the democratization of analysis of complex information.

For those interested in expanding the capabilities of the system, there are multiple promising directions for its evolution:

• Incorporation of automatic data verification mechanisms to ensure accuracy.
• Implementation of multimodal capabilities that allow incorporating images and visualizations.
• Integration with more sources of information and knowledge bases.
• Development of more intuitive user interfaces for human intervention.
• Expansion to specialized domains such as medicine, law or sciences.

In summary, this exercise not only demonstrates the feasibility of automating the generation of complex reports through artificial intelligence, but also points to a promising path towards a future where deep analysis of any topic is within everyone's reach, regardless of their level of technical experience. The combination of advanced language models, graph architectures and parallelization techniques opens a range of possibilities to transform the way we generate and consume information.

 The Spanish Data Protection Agency  has recently published the Spanish translation of the Guide on Synthetic Data Generation, originally produced by the Data Protection Authority of Singapore. This document provides technical and practical guidance for data protection officers, managers and data protection officers on how to implement this technology that allows simulating real data while maintaining their statistical characteristics without compromising personal information.

The guide highlights how synthetic data can drive the data economy, accelerate innovation and mitigate risks in security breaches. To this end, it presents case studies, recommendations and best practices aimed at reducing the risks of re-identification. In this post, we analyse the key aspects of the Guide highlighting main use cases and examples of practical application.

What are synthetic data? Concept and benefits

Synthetic data is artificial data generated using mathematical models specifically designed for artificial intelligence (AI) or machine learning (ML) systems. This data is created by training a model on a source dataset to imitate its characteristics and structure, but without exactly replicating the original records.

High-quality synthetic data retain the statistical properties and patterns of the original data. They therefore allow for analyses that produce results similar to those that would be obtained with real data. However, being artificial, they significantly reduce the risks associated with the exposure of sensitive or personal information.

For more information on this topic, you can read this Monographic report on synthetic data:. What are they and what are they used for? with detailed information on the theoretical foundations, methodologies and practical applications of this technology.

The implementation of synthetic data offers multiple advantages for organisations, for example:

• Privacy protection: allow data analysis while maintaining the confidentiality of personal or commercially sensitive information.
• Regulatory compliance: make it easier to follow data protection regulations while maximising the value of information assets.
• Risk reduction: minimise the chances of data breaches and their consequences.
• Driving innovation: accelerate the development of data-driven solutions without compromising privacy.
• Enhanced collaboration: Enable valuable information to be shared securely across organisations and departments.

Steps to generate synthetic data

To properly implement this technology, the Guide on Synthetic Data Generation recommends following a structured five-step approach:

1. Know the data: cClearly understand the purpose of the synthetic data and the characteristics of the source data to be preserved, setting precise targets for the threshold of acceptable risk and expected utility.
2. Prepare the data: iidentify key insights to be retained, select relevant attributes, remove or pseudonymise direct identifiers, and standardise formats and structures in a well-documented data dictionary .
3. Generate synthetic data: sselect the most appropriate methods according to the use case, assess quality through completeness, fidelity and usability checks, and iteratively adjust the process to achieve the desired balance.
4. Assess re-identification risks: aApply attack-based techniques to determine the possibility of inferring information about individuals or their membership of the original set, ensuring that risk levels are acceptable.
5. Manage residual risks: iImplement technical, governance and contractual controls to mitigate identified risks, properly documenting the entire process.

Practical applications and success stories

To realise all these benefits, synthetic data can be applied in a variety of scenarios that respond to specific organisational needs. The Guide mentions, for example:

1 Generation of datasets for training AI/ML models: lSynthetic data solves the problem of the scarcity of labelled (i.e. usable) data for training AI models. Where real data are limited, synthetic data can be a cost-effective alternative. In addition, they allow to simulate extraordinary events or to increase the representation of minority groups in training sets. An interesting application to improve the performance and representativeness of all social groups in AI models.

2 Data analysis and collaboration: eThis type of data facilitates the exchange of information for analysis, especially in sectors such as health, where the original data is particularly sensitive. In this sector as in others, they provide stakeholders with a representative sample of actual data without exposing confidential information, allowing them to assess the quality and potential of the data before formal agreements are made.

3 Software testing: sis very useful for system development and software testing because it allows the use of realistic, but not real data in development environments, thus avoiding possible personal data breaches in case of compromise of the development environment..

The practical application of synthetic data is already showing positive results in various sectors:

I. Financial sector: fraud detection. J.P. Morgan has successfully used synthetic data to train fraud detection models, creating datasets with a higher percentage of fraudulent cases that significantly improved the models' ability to identify anomalous behaviour.

II. Technology sector: research on AI bias. Mastercard collaborated with researchers to develop methods to test for bias in AI using synthetic data that maintained the true relationships of the original data, but were private enough to be shared with outside researchers, enabling advances that would not have been possible without this technology.

III. Health sector: safeguarding patient data. Johnson & Johnson implemented AI-generated synthetic data as an alternative to traditional anonymisation techniques to process healthcare data, achieving a significant improvement in the quality of analysis by effectively representing the target population while protecting patients' privacy.

The balance between utility and protection

It is important to note that synthetic data are not inherently risk-free. The similarity to the original data could, in certain circumstances, allow information about individuals or sensitive data to be leaked. It is therefore crucial to strike a balance between data utility and data protection.

This balance can be achieved by implementing good practices during the process of generating synthetic data, incorporating protective measures such as:

• Adequate data preparation: removal of outliers, pseudonymisation of direct identifiers and generalisation of granular data.
• Re-identification risk assessment: analysis of the possibility that synthetic data can be linked to real individuals.
• Implementation of technical controls: adding noise to data, reducing granularity or applying differential privacy techniques.

Synthetic data represents a exceptional opportunity to drive data-driven innovation while respecting privacy and complying with data protection regulations. Their ability to generate statistically representative but artificial information makes them a versatile tool for multiple applications, from AI model training to inter-organisational collaboration and software development.

By properly implementing the best practices and controls described in Guide on synthetic data generation translated by the AEPD, organisations can reap the benefits of synthetic data while minimising the associated risks, positioning themselves at the forefront of responsible digital transformation. The adoption of privacy-enhancing technologies such as synthetic data is not only a defensive measure, but a proactive step towards an organisational culture that values both innovation and data protection, which are critical to success in the digital economy of the future.

The evolution of generative AI has been dizzying: from the first great language models that impressed us with their ability to reproduce human reading and writing, through the advanced RAG (Retrieval-Augmented Generation) techniques that quantitatively improved the quality of the responses provided and the emergence of intelligent agents, to an innovation that redefines our relationship with technology: Computer use.

At the end of April 2020, just one month after the start of an unprecedented period of worldwide home confinement due to the SAR-Covid19 global pandemic, we spread from datos.gob.es the large GPT-2 and GPT-3 language models. OpenAI, founded in 2015, had presented almost a year earlier (February 2019) a new language model that was able to generate written text virtually indistinguishable from that created by a human. GPT-2 had been trained on a corpus (set of texts prepared to train language models) of about 40 GB (Gigabytes) in size (about 8 million web pages), while the latest family of models based on GPT-4 is estimated to have been trained on TB (Terabyte) sized corpora; a thousand times more.

In this context, it is important to talk about two concepts:

• LLLMs (Large Language Models ): are large-scale language models, trained on vast amounts of data and capable of performing a wide range of linguistic tasks. Today, we have countless tools based on these LLMs that, by field of expertise, are able to generate programming code, ultra-realistic images and videos, and solve complex mathematical problems. All major companies and organisations in the digital-technology sector have embarked on integrating these tools into their different software and hardware products, developing use cases that solve or optimise specific tasks and activities that previously required a high degree of human intervention.

• Agents: The user experience with artificial intelligence models is becoming more and more complete, so that we can ask our interface not only to answer our questions, but also to perform complex tasks that require integration with other IT tools. For example, we not only ask a chatbot for information on the best restaurants in the area, but we also ask them to search for table availability for specific dates and make a reservation for us. This extended user experience is what artificial intelligence agentsprovide us with. Based on the large language models, these agents are able to interact with the outside world (to the model) and "talk" to other services via APIs and programming interfaces prepared for this purpose.

Computer use

However, the ability of agents to perform actions autonomously depends on two key elements: on the one hand, their concrete programming - the functionality that has been programmed or configured for them; on the other hand, the need for all other programmes to be ready to "talk" to these agents. That is, their programming interfaces must be ready to receive instructions from these agents. For example, the restaurant reservation application has to be prepared, not only to receive forms filled in by a human, but also requests made by an agent that has been previously invoked by a human using natural language. This fact imposes a limitation on the set of activities and/or tasks that we can automate from a conversational interface. In other words, the conversational interface can provide us with almost infinite answers to the questions we ask it, but it is severely limited in its ability to interact with the outside world due to the lack of preparation of the rest of computer applications.

This is where Computer use comes in. With the arrival of the Claude 3.5 Sonnet model, Anthropic has introduced Computer use, a beta capability that allows AI to interact directly with graphical user interfaces.

How does Computer use work?

Claude can move your computer cursor as if it were you, click buttons and type text, emulating the way humans operate a computer. The best way to understand how Computer use works in practice is to see it in action. Here is a link directly to the YouTube channel of the specific Computer use section.

Figure 1. Screenshot from Anthropic's YouTube channel, Computer use specific section.

Would you like to try it?

If you've made it this far, you can't miss out without trying it with your own hands.

Here is a simple guide to testing Computer use in an isolated environment. It is important to take into account the security recommendations that Antrophic proposes in its Computer use guidelines. This feature of the Claude Sonet model can perform actions on a computer and this can be potentially dangerous, so it is recommended to carefully review the security warning of Computer use.

All official developer documentation can be found in the antrophic's official Github repository. In this post, we have chosen to run Computer use in a Docker container environment. It is the easiest and safest way to test it. If you don't already have it, you can follow the simple official guidelines to pre-install it on your system.

To reproduce this test we propose to follow this script step by step:.

1. Anthropic API Key. To interact with Claude Sonet you need an Anthropic account which you can create for free here. Once inside, you can go to the API Keys section and create a new one for your test

2. Once you have your API Key, you must run this command in your terminal, substituting your key where it says "%your_api_key%":

The enormous acceleration of innovation in artificial intelligence (AI) in recent years has largely revolved around the development of so-called "foundational models". Also known as Large [X] Models (Large [X] Models or LxM), Foundation Models, as defined by the Center for Research on Foundation Models (CRFM) of the Institute for Human-Centered Artificial Intelligence's (HAI) Stanford University's models that have been trained on large and highly diverse datasets and can be adapted to perform a wide range of tasks using techniques such as fine-tuning (fine-tuning).

It is precisely this versatility and adaptability that has made foundational models the cornerstone of the numerous applications of artificial intelligence being developed, since a single base architecture can be used across a multitude of use cases with limited additional effort.

Types of foundational models

The "X" in LxM can be replaced by several options depending on the type of data or tasks for which the model is specialised. The best known by the public are the LLM (Large Language Models), which are at the basis of applications such as ChatGPT or Gemini, and which focus on natural language understanding and generation.. LVMs (Large Vision Models), such as DINOv2 or CLIP, are designed tointerpret images and videos, recognise objects or generate visual descriptions..  There are also models such as Operator or Rabbit R1 that fall into the LAM (Large Action Models) category and are aimed atexecuting actions from complex instructions..

As regulations have emerged in different parts of the world, so have other definitions that seek to establish criteria and responsibilities for these models to foster confidence and security. The most relevant definition for our context is that set out in the European Union AI Regulation (AI Act), which calls them "general-purpose AI models" and distinguishes them by their "ability to competently perform a wide variety of discrete tasks" and because they are "typically trained using large volumes of data and through a variety of methods, such as self-supervised, unsupervised or reinforcement learning".

Foundational models in Spanish and other co-official languages

Historically, English has been the dominant language in the development of large AI models, to the extent that around 90% of the training tokens of today's large models are drawn from English texts. It is therefore logical that the most popular models, for example OpenAI's GPT family, Google's Gemini or Meta's Llama, are more competent at responding in English and perform less well when used in other languages such as Spanish.

Therefore, the creation of foundational models in Spanish, such as ALIA, is not simply a technical or research exercise, but a strategic move to ensure that artificial intelligence does not further deepen the linguistic and cultural asymmetries that already exist in digital technologies in general. The development of ALIA, driven by the Spain's Artificial Intelligence Strategy 2024, "based on the broad scope of our languages, spoken by 600 million people, aims to facilitate the development of advanced services and products in language technologies, offering an infrastructure marked by maximum transparency and openness".

Such initiatives are not unique to Spain. Other similar projects include BLOOM, a 176-billion-parameter multilingual model developed by more than 1,000 researchers worldwide and supporting 46 natural languages and 13 programming languages. In China, Baidu has developed ERNIE, a model with strong Mandarin capabilities, while in France the PAGNOL model has focused on improving French capabilities. These parallel efforts show a global trend towards the "linguistic democratisation" of AI.

 Since the beginning of 2025, the first language models in the four co-official languages, within the ALIA project, have been available.  The ALIA family of models includes ALIA-40B, a model with 40.40 billion parameters, which is currently the most advanced public multilingual foundational model in Europeand which was trained for more than 8 months on the MareNostrum 5 supercomputer, processing 6.9 trillion tokens equivalent to about 33 terabytes of text (about 17 million books!). Here all kinds of official documents and scientific repositories in Spanish are included, from congressional journals to scientific repositories or official bulletins to ensure the richness and quality of your knowledge.

Although this is a multilingual model, Spanish and co-official languages have a much higher weight than usual in these models, around 20%, as the training of the model was designed specifically for these languages, reducing the relevance of English and adapting the tokens to the needs of Spanish, Catalan, Basque and Galician.. As a result, ALIA "understands" our local expressions and cultural nuances better than a generic model trained mostly in English.

Applications of the foundational models in Spanish and co-official languages

It is still too early to judge the impact on specific sectors and applications that ALIA and other models that may be developed from this experience may have. However, they are expected to serve as a basis for improving many Artificial Intelligence applications and solutions:.

• Public administration and government: ALIA could give life to virtual assistants that attend to citizens 24 hours a day for procedures such as paying taxes, renewing ID cards, applying for grants, etc., as it is specifically trained in Spanish regulations.  In fact, a pilot for the Tax Agency using ALIA, which would aim to streamline internal procedures, has already been announced.
• Education: A model such as ALIA could also be the basis for personalised virtual tutors to guide students in Spanish and co-official languages. For example, assistants who explain concepts of mathematics or history in simple language and answer questions from the students, adapting to their level since, knowing our language well, they would be able to provide important nuances in the answers and understand the typical doubts of native speakers in these languages. They could also help teachers by generating exercises or summaries of readings or assisting them in correcting students' work.
• Health: ALIA could be used to analyse medical texts and assist healthcare professionals with clinical reports, medical records, information leaflets, etc. For example, it could review patient files to extract key elements, or assist professionals in the diagnostic process.  In fact, the Ministry of Health is planning a pilot application with ALIA to improve early detection of heart failure in primary care.
• Justice: In the legal field, ALIA would understand technical terms and contexts of Spanish law much better than a non-specialised model as it has been trained with legal vocabulary from official documents. An ALIA-based virtual paralegal could be able to answer basic citizen queries, such as how to initiate a given legal procedure, citing the applicable law. The administration of justice could also benefit from much more accurate machine translations of court documents between co-official languages.

Future lines

The development of foundation models in Spanish, as in other languages, is beginning to be seen outside the United States as a strategic issue that contributes to guaranteeing the technological sovereignty of countries.  Of course, it will be necessary to continue training more advanced versions (models with up to 175 billion parameters are targeted, which would be comparable to the most powerful in the world), incorporating new open data, and fine-tuning applications. From the Data Directorate and the SEDIA it is intended to continue to support the growth of this family of models, to keep it at the forefront and ensure its adoption.

On the other hand, these first foundational models in Spanish and co-official languages have initially focused on written language, so the next natural frontier could be multimodality. Integrating the capacity to manage images, audio or video in Spanish together with the text would multiply its practical applications, since the interpretation of images in Spanish is one of the areas where the greatest deficiencies are detected in the large generic models.

Ethical issues will also need to be monitored to ensure that these models do not perpetuate bias and are useful for all groups, including those who speak different languages or have different levels of education.   In this respect,  Explainable Artificial Intelligence (XAI) is not optional, but a fundamental requirement to ensure its responsible adoption..  The National AI Supervisory Agency, the research community and civil society itself will have an important role to play here.

Content prepared by Jose Luis Marín, Senior Consultant in Data, Strategy, Innovation & Digitalization. The contents and views reflected in this publication are the sole responsibility of the author.

Did you know that data science skills are among the most in-demand skills in business? In this podcast, we are going to tell you how you can train yourself in this field, in a self-taught way. For this purpose, we will have two experts in data science:

• Juan Benavente, industrial and computer engineer with more than 12 years of experience in technological innovation and digital transformation. In addition, it has been training new professionals in technology schools, business schools and universities for years.
• Alejandro Alija, PhD in physics, data scientist and expert in digital transformation.  In addition to his extensive professional experience focused on the Internet of Things (internet of things), Alejandro also works as a lecturer in different business schools and universities.

Listen to the podcast (in spanish)

Summary of the interview

1. What is data science? Why is it important and what can it do for us? 

Alejandro Alija: Data science could be defined as a discipline whose main objective is to understand the world, the processes of business and life, by analysing and observing data.Data science is a discipline whose main objective is to understand the world, the processes of business and life, by analysing and observing the data.. In the last 20 years it has gained exceptional relevance due to the explosion in data generation, mainly due to the irruption of the internet and the connected world.

Juan Benavente:  The term data science has evolved since its inception. Today, a data scientist is the person who is working at the highest level in data analysis, often associated with the building of machine learning or artificial intelligence algorithms for specific companies or sectors, such as predicting or optimising manufacturing in a plant.

The profession is evolving rapidly, and is likely to fragment in the coming years. We have seen the emergence of new roles such as data engineers or MLOps specialists. The important thing is that today any professional, regardless of their field, needs to work with data. There is no doubt that any position or company requires increasingly advanced data analysis. It doesn't matter if you are in marketing, sales, operations or at university. Anyone today is working with, manipulating and analysing data. If we also aspire to data science, which would be the highest level of expertise, we will be in a very beneficial position. But I would definitely recommend any professional to keep this on their radar.

2. How did you get started in data science and what do you do to keep up to date? What strategies would you recommend for both beginners and more experienced profiles?

Alejandro Alija: My basic background is in physics, and I did my PhD in basic science. In fact, it could be said that any scientist, by definition, is a data scientist, because science is based on formulating hypotheses and proving them with experiments and theories. My relationship with data started early in academia. A turning point in my career was when I started working in the private sector, specifically in an environmental management company that measures and monitors air pollution. The environment is a field that is traditionally a major generator of data, especially as it is a regulated sector where administrations and private companies are obliged, for example, to record air pollution levels under certain conditions. I found historical series up to 20 years old that were available for me to analyse. From there my curiosity began and I specialised in concrete tools to analyse and understand what is happening in the world.

Juan Benavente: I can identify with what Alejandro said because I am not a computer scientist either. I trained in industrial engineering and although computer science is one of my interests, it was not my base. In contrast, nowadays, I do see that more specialists are being trained at the university level.  A data scientist today has manyskills on their back such as statistics, mathematics and the ability to understand everything that goes on in the industry. I have been acquiring this knowledge through practice. On how to keep up to date, I think that, in many cases, you can be in contact with companies that are innovating in this field. A lot can also be learned at industry or technology events. I started in the smart cities and have moved on to the industrial world to learn little by little.

Alejandro Alija:. To add another source to keep up to date. Apart from what Juan has said, I think it's important to identify what we call outsiders, the manufacturers of technologies, the market players.  They are a very useful source of information to stay up to date: identify their futures strategies and what they are betting on.

3. If someone with little or no technical knowledge wants to learn data science, where do they start?

Juan Benavente:  In training, I have come across very different profiless: from people who have just graduated from university to profiles that have been trained in very different fields and find in data science an opportunity to transform themselves and dedicate themselves to this.  Thinking of someone who is just starting out, I think the best thing to do is put your knowledge into practice. In projects I have worked on, we defined the methodology in three phases: a first phase of more theoretical aspects, taking into account mathematics, programming and everything a data scientist needs to know; once you have those basics, the sooner you start working and practising those skills, the better. I believe that skill sharpens the wit and, both to keep up to date and to train yourself and acquire useful knowledge, the sooner you enter into a project, the better. And even more so in a world that is so frequently updated. In recent years, the emergence of the Generative AI has brought other opportunities.  There are also opportunities for new profiles who want to be trained . Even if you are not an expert in programming, you have tools that can help you with programming, and the same can happen in mathematics or statistics.

Alejandro Alija:. To complement what Juan says from a different perspective. I think it is worth highlighting the evolution of the data science profession.. I remember when that paper about "the sexiest profession in the world" became famous and went viral, but then things adjusted. The first settlers in the world of data science did not come so much from computer science or informatics. There were more outsiders: physicists, mathematicians, with a strong background in mathematics and physics, and even some engineers whose work and professional development meant that they ended up using many tools from the computer science field. Gradually, it has become more and more balanced. It is now a discipline that continues to have those two strands: people who come from the world of physics and mathematics towards the more basic data, and people who come with programming skills. Everyone knows what they have to balance in their toolbox. Thinking about a junior profile who is just starting out, I think a very important thing - and we see this when we teach - is programming skills. I would say that having programming skills is not just a plus, but a basic requirement for advancement in this profession. It is true that some people can do well without a lot of programming skills, but I would argue that a beginner needs to have those first programming skills with a basic toolset . We're talking about languages such as Python and R, which are the headline languages. You don't need to be a great coder, but you do need to have some basic knowledge to get started. Then, of course, specific training in the mathematical foundations of data science is crucial. The fundamental statistics and more advanced statistics are complements that, if present, will move a person along the data science learning curve much faster. Thirdly, I would say that specialisation in particular tools is important. Some people are more oriented towards data engineering, others towards the modelling world. Ideally, specialise in a few frameworks and use them together, as optimally as possible.

4. In addition to teaching, you both work in technology companies. What technical certifications are most valued in the business sector and what open sources of knowledge do you recommend to prepare for them?

Juan Benavente: Personally, it's not what I look at most, but I think it can be relevant, especially for people who are starting out and need help in structuring their approach to the problem and understanding it. I recommend certifications of technologies that are in use in any company where you want to end up working. Especially from providers of cloud computing and widespread data analytics tools. These are certifications that I would recommend for someone who wants to approach this world and needs a structure to help them. When you don't have a knowledge base, it can be a bit confusing to understand where to start. Perhaps you should reinforce programming or mathematical knowledge first, but it can all seem a bit complicated. Where these certifications certainly help you is, in addition to reinforcing concepts, to ensure that you are moving well and know the typical ecosystem of tools you will be working with tomorrow. It is not just about theoretical concepts, but about knowing the ecosystems that you will encounter when you start working, whether you are starting your own company or working in an established company. It makes it much easier for you to get to know the typical ecosystem of tools. Call it Microsoft Computing, Amazon or other providers of such solutions. This will allow you to focus more quickly on the work itself, and less on all the tools that surround it. I believe that this type of certification is useful, especially for profiles that are approaching this world with enthusiasm. It will help them both to structure themselves and to land well in their professional destination. They are also likely to be valued in selection processes.

Alejandro Alija: If someone listens to us and wants more specific guidelines, it could be structured in blocks. There are a series of massive online courses that, for me, were a turning point. In my early days, I tried to enrol in several of these courses on platforms such as Coursera, edX, where even the technology manufacturers themselves design these courses. I believe that this kind of massive, self-service, online courses provide a good starting base. A second block would be the courses and certifications of the big technology providers, such as Microsoft, Amazon Web Services, Google and other platforms that are benchmarks in the world of data. These companies have the advantage that their learning paths are very well structured, which facilitates professional growth within their own ecosystems. Certifications from different suppliers can be combined. For a person who wants to go into this field, the path ranges from the simplest to the most advanced certifications, such as being a data solutions architect or a specialist in a specific data analytics service or product. These two learning blocks are available on the internet, most of them are open and free or close to free. Beyond knowledge, what is valued is certification, especially in companies looking for these professional profiles.

5. In addition to theoretical training, practice is key, and one of the most interesting methods of learning is to replicate exercises step by step. In this sense, from datos.gob.es we offer didactic resources, many of them developed by you as experts in the project, can you tell us what these exercises consist of?. How are they approached?

Alejandro Alija: The approach we always took was designed for a broad audience, without complex prerequisites. We wanted any user of the portal to be able to replicate the exercises, although it is clear that the more knowledge you have, the more you can use it to your advantage. Exercises have a well-defined structure: a documentary section, usually a content post or a report describing what the exercise consists of, what materials are needed, what the objectives are and what it is intended to achieve. In addition, we accompany each exercise with two additional resources. The first resource is a code repository where we upload the necessary materials, with a brief description and the code of the exercise. It can be a Python notebook , a Jupyter Notebook or a simple script, where the technical content is. And then another fundamental element that we believe is important and that is aimed at facilitating the execution of the exercises. In data science and programming, non-specialist users often find it difficult to set up a working environment. A Python exercise, for example, requires having a programming environment installed, knowing the necessary libraries and making configurations that are trivial for professionals, but can be very complex for beginners. To mitigate this barrier, we publish most of our exercises on Google Colab, a wonderful and open tool. Google Colab is a web programming environment where the user only needs a browser to access it. Basically, Google provides us with a virtual computer where we can run our programmes and exercises without the need for special configurations. The important thing is that the exercise is ready to use and we always check it in this environment, which makes it much easier to learn for beginners or less technically experienced users.

Juan Benavente: Yes, we always take a user-oriented approach, step by step, trying to make it open and accessible. The aim is for anyone to be able to run an exercise without the need for complex configurations, focusing on topics as close to reality as possible. We often take advantage of open data published by entities such as the DGT or other bodies to make realistic analyses. We have developed very interesting exercises, such as energy market predictions, analysis of critical materials for batteries and electronics, which allow learning not only about technology, but also about the specific subject matter.. You can get down to work right away, not only to learn, but also to find out about the subject.

6. In closing, we'd like you to offer a piece of advice that is more attitude-oriented than technical, what would you say to someone starting out in data science?

Alejandro Alija:  As for an attitude tip for someone starting out in data science, I suggest be brave. There is no need to worry about being unprepared, because in this field everything is to be done and anyone can contribute value. Data science is multi-faceted: there are professionals closer to the business world who can provide valuable insights, and others who are more technical and need to understand the context of each area. My advice is to be content with the resources available without panicking, because, although the path may seem complex, the opportunities are very high. As a technical tip, it is important to be sensitive to the development and use of data. The more understanding one has of this world, the smoother the approach to projects will be.

Juan Benavente: I endorse the advice to be brave and add a reflection on programming: many people find the theoretical concept attractive, but when they get to practice and see the complexity of programming, some are discouraged by lack of prior knowledge or different expectations. It is important to add the concepts of patience and perseverance. When you start in this field, you are faced with multiple areas that you need to master: programming, statistics, mathematics, and specific knowledge of the sector you will be working in, be it marketing, logistics or another field. The expectation of becoming an expert quickly is unrealistic. It is a profession that, although it can be started without fear and by collaborating with professionals, requires a journey and a learning process. You have to be consistent and patient, managing expectations appropriately. Most people who have been in this world for a long time agree that they have no regrets about going into data science. It is a very attractive profession where you can add significant value, with an important technological component. However, the path is not always straightforward. There will be complex projects, moments of frustration when analyses do not yield the expected results or when working with data proves more challenging than expected. But looking back, few professionals regret having invested time and effort in training and developing in this field. In summary, the key tips are: courage to start, perseverance in learning and development of programming skills.

The European Union is at the forefront of the development of safe, ethical and people-centred artificial intelligence (AI). Through a robust regulatory framework, based on human rights and fundamental values, the EU is building an AI ecosystem that simultaneously benefits citizens, businesses and public administrations.  As part of its commitment to the proper development of this technology, the European Commission has proposed a set of actions to promote its excellence.

In this regard, a pioneering piece of legislation that establishes a comprehensive legal framework stands out: the AI Act.  It classifies artificial intelligence models according to their level of risk and establishes specific obligations for providers regarding data and data governance. In parallel, the Coordinated Plan on AI updated in 2021 sets out a roadmap to boost investment, harmonise policies and encourage the uptake of AI across the EU.

 Spain is aligned with Europe in this area and therefore has a strategy to accelerate its development and expansion.. In addition, the transposition of the AI law has recently been approved, with the preliminary draft law for an ethical, inclusive and beneficial use of artificial intelligence.

European projects transforming key sectors

In this context, the EU is funding numerous projects that use artificial intelligence technologies to solve challenges in various fields. Below, we highlight some of the most innovative ones, some of which have already been completed and some of which are underway:

Agriculture and food sustainability

Projects currently underway:

• ANTARES: develops smart sensor technologies and big data to help farmers produce more food in a sustainable way, benefiting society, farm incomes and the environment.

Examples of other completed projects:

• Pantheon: developed a control and data acquisition system, equivalent to industrial SCADA, for precision farming in large hazelnut orchards, increasing production, reducing chemical inputs and simplifying management.

• Trimbot2020: researched robotics and vision technologies to create the first outdoor gardening robot, capable of navigating varied terrain and trimming rose bushes, hedges and topiary.

Industry and manufacturing

Projects currently underway:

• SERENA: applies AI techniques to predict maintenance needs of industrial equipment, reducing costs and time, and improving the productivity of production processes..

• SecondHands: has developed a robot capable of proactively assisting maintenance technicians by recognising human activity and anticipating their needs, increasing efficiency and productivity in industrial environments.

Examples of other completed projects:

• QU4LITY: combined data and AI to increase manufacturing sustainability, providing a data-shared, SME-friendly, standardised and transformative zero-defect manufacturing model.

• KYKLOS 4.0: explored how cyber-physical systems, product lifecycle management, augmented reality and AI can transform circular manufacturing through seven large-scale pilot projects.

Transport and mobility

Projects currently underway

• VI-DAS: A project by a Spanish company working on advanced driver assistance systems and navigation aids, combining traffic understanding with consideration of the driver's physical, mental and behavioural state to improve road safety.

• PILOTING: adapts, integrates and demonstrates robotic solutions in an integrated platform for the inspection and maintenance of refineries, bridges and tunnels.. One of its focuses is on boosting production and access to inspection data.

Examples of other completed projects:

• FABULOS: has developed and tested a local public transport system using autonomous minibuses, demonstrating its viability and promoting the introduction of robotic technologies in public infrastructure.

Social impact research

Projects currently underway:

• HUMAINT: provides a multidisciplinary understanding of the current state and future evolution of machine intelligence and its potential impact on human behaviour, focusing on cognitive and socio-emotional capabilities.

• AI Watch: monitors industrial, technological and research capacity, policy initiatives in Member States, AI adoption and technical developments, and their impact on the economy, society and public services.

Examples of other completed projects:

• TECHNEQUALITY: examined the potential social consequences of the digital age, looking at how AI and robots affect work and how automation may impact various social groups differently.

Health and well-being

Projects currently underway:

• DeepHealth: develops advanced tools for medical image processing and predictive modelling, facilitating the daily work of healthcare personnel without the need to combine multiple tools..

• BigO: collects and analyses anonymised data on child behaviour patterns and their environment to extract evidence on local factors involved in childhood obesity.

Examples of other completed projects:

• PRIMAGE: has created a cloud-based platform to support decision making for malignant solid tumours, offering predictive tools for diagnosis, prognosis and monitoring, using imaging biomarkers and simulation of tumour growth..

• SelfBACK: provided personalised support to patients with low back pain through a mobile app, using sensor-collected data to tailor recommendations to each user.

• EYE-RISK: developed tools that predict the likelihood of developing age-related eye diseases and measures to reduce this risk, including a diagnostic panel to assess genetic predisposition.

• Solve-RD: improved diagnosis of rare diseases by pooling patient data and advanced genetic methods.

The future of AI in Europe

These examples, both past and present, are very interesting use cases of the development of artificial intelligence in Europe. However, the EU's commitment to AI is also forward-looking. And it is reflected in an ambitious investment plan: the Commission plans to invest EUR 1 billion per year in AI, from the Digital Europe and Horizon Europe programmes, with the aim of attracting more than EUR 20 billion of total AI investment per year during this decade..

The development of an ethical, transparent and people-centred IA is already an EU objective that goes beyond the legal framework. With a hands-on approach, the European Union funds projects that not only drive technological innovation, but also address key societal challenges, from health to climate change, building a more sustainable, inclusive and prosperous future for all European citizens.

There is no doubt that artificial intelligence has become a fundamental pillar of technological innovation.  Today, artificial intelligence (AI) can create chatbots specialised in open data, applications that facilitate professional work and even a digital Earth model to anticipate natural disasters.

The possibilities are endless, however, the future of AI also has challenges to overcome to make models more inclusive, accessible and transparent. In this respect, the European Union is developing various initiatives to make progress in this field.

European regulatory framework for a more open and transparent AI.

The EU's approach to AI seeks to give citizens the confidence to adopt these technologies and to encourage businesses to develop them. To this end, the European AI Regulation sets out guidelines for the development of artificial intelligence in line with European values of privacy, security and cultural diversity. On the other hand, the Data Governance Regulation (DGA) defines that broad access to data must be guaranteed without compromising intellectual property rights, privacy and fairness.

Together with the Artificial Intelligence Act, the update of the Coordinated Plan on AI ensures the security and fundamental rights of individuals and businesses, while strengthening investment and innovation in all EU countries. The Commission has also launched an Artificial Intelligence Innovation Package to help European start-ups and SMEs develop reliable AI that respects EU values and standards.

Other institutions are also working on boosting intelligence by pushing open source AI models as a very interesting solution. A recent report by Open Future and Open Source Initiative (OSI) defines what data governance should look like in open source AI models. One of the challenges highlighted in the report is precisely to strike a balance between open data and data rights, to achieve more transparency and to avoid cultural bias. In fact, experts in the field Ricard Martínez and Carmen Torrijos debated this issue in the pódcast of datos.gob.es.

The OpenEuroLLM project

With the aim of solving potential challenges and as an innovative and open solution, the European Union, through the Digital Europe programme has presented  through this open source artificial intelligence project it is expected to create efficient, transparent language models aligned with European AI regulations.

The OpenEuroLLM project has as its main goal the development of state-of-the-art language models for a wide variety of public and private applications. Among the most important objectives, we can mention the following:

1. Extend the multilingual capabilities of existing models: this includes not only the official languages of the European Union, but also other languages that are of social and economic interest. Europe is a continent rich in linguistic diversity, and the project seeks to reflect this diversity in AI models.
2. Sustainable access to fundamental models: lthe models developed within the project will be easy to access and ready to be adjusted to various applications. This will not only benefit large enterprises, but also small and medium-sized enterprises (SMEs) that wish to integrate AI into their processes without facing technological barriers.
3. Evaluation of results and alignment with European regulations: models will be evaluated according to rigorous safety standards and alignment with the European AI Regulation and other European regulatory frameworks. This will ensure that AI solutions are safe and respect fundamental rights.
4. Transparency and accessibility: One of the premises of the project is to openly share the tools, processes and intermediate results of the training processes. This will allow other researchers and developers to reproduce, improve and adapt the models for their own purposes.
5. Community building: OpenEuroLLM is not limited to modelling but also aims to build an active and engaged community, both in the public and private sector, that can collaborate, share knowledge and work together to advance AI research.

The OpenEuroLLM Consortium: a collaborative and multinational project

The OpenEuroLLM project is being developed by a consortium of 20 European research institutions , technology companies and supercomputing centres, under the coordination of Charles University (Czech Republic) and the collaboration of Silo GenAI (Finland). The consortium brings together some of the leading institutions and companies in the field of artificial intelligence in Europe, creating a multinational collaboration to develop open source language models.

The main institutions participating in the project include renowned universities such as University of Helsinki (Finland) and University of Oslo (Norway), as well as technology companies such as Aleph Alpha Research (Germany) or the company from Elche prompsit (Spain), among others. In addition, supercomputing centres such as the Barcelona Supercomputing Center (Spain) or SURF (The Netherlands) provide the infrastructure needed to train large-scale models.

Linguistic diversity, transparency and compliance with EU standards

One of the biggest challenges of globalised artificial intelligence is the inclusion of multiple languages and the preservation of cultural differences. Europe, with its vast linguistic diversity, presents a unique environment in which to address these issues. OpenEuroLLM is committed to preserving this diversity and ensuring that the AI models developed are sensitive to the linguistic and cultural variations of the region.

As we saw at the beginning of this post, technological development must go hand in hand with ethical and responsible values. In this respect, one of the key features of the OpenEuroLLM project is its focus on transparency. Models, data, documentation, training code and evaluation metrics will be fully available to the public. This will allow researchers and developers to audit, modify and improve the models, ensuring an open and collaborative approach.

In addition, the project is aligned with strict European AI regulations. OpenEuroLLM is designed to comply with the EU's AI Law , which sets stringent criteria to ensure safety, fairness and privacy in artificial intelligence systems.

Democratising access to AI

One of the most important achievements of OpenEuroLLLM is the democratisation of access to high-performance AI. Open source models will enable businesses, academic institutions and public sector organisations across Europe to have access to cutting-edge technology, regardless of their size or budget.

This is especially relevant for small and medium-sized enterprises (SMEs), which often face difficulties in accessing AI solutions due to high licensing costs or technological barriers. OpenEuroLLM will remove these barriers and enable companies to develop innovative products and services using AI, which will contribute to Europe's economic growth.

The OpenEuroLLM project is also an EU commitment to digital sovereignty that is strategically investing in the development of technological infrastructure that reduces dependence on global players and strengthens European competitiveness in the field of artificial intelligence. This is an important step towards artificial intelligence that is not only more advanced, but also fairer, safer and more responsible.

The increasing adoption of artificial intelligence (AI) systems in critical areas such as public administration, financial services or healthcare has brought the need for algorithmic transparency to the forefront. The complexity of AI models used to make decisions such as granting credit or making a medical diagnosis, especially when it comes to deep learning algorithms, often gives rise to what is commonly referred to as the "black box" problem, i.e. the difficulty of interpreting and understanding how and why an AI model arrives at a certain conclusion. The LLLMs or SLMs that we use so much lately are a clear example of a black box system where not even the developers themselves are able to foresee their behaviour.

In regulated sectors, such as finance or healthcare, AI-based decisions can significantly affect people's lives and therefore it is not acceptable to raise doubts about possible bias or attribution of responsibility. As a result, governments have begun to develop regulatory frameworks such as the Artificial Intelligence Regulation that require greater explainability and oversight in the use of these systems with the additional aim of generating confidence in the advances of the digital economy.

Explainable artificial intelligence (XAI) is the discipline that has emerged in response to this challenge, proposing methods to make the decisions of AI models understandable. As in other areas related to artificial intelligence, such as LLLM training, open data is an important ally of explainable artificial intelligence to build audit and verification mechanisms for algorithms and their decisions.

What is explainable AI (XAI)?

Explainable AI refers to methods and tools that allow humans to understand and trust the results of machine learning models. According to the U.S. National Institute of Standards and Technology (NIST), the NIST is the only organisation in the U.S. that has a national standards body. The four key principles of Explainable Artificial Intelligence in the US are to ensure that AI systems are transparent, understandable and trusted by users:

• Explainability (Explainability): the AI must provide clear and understandable explanations of how it arrives at its decisions and recommendations.
• Meaningful (Meaningful): explanations must be meaningful and understandable to users.
• Accuracy (Accuracy): AI must generate accurate and reliable results, and the explanation of these results must accurately reflect its performance.
• Knowledge Limits (Knowledge Limits): AI must recognise when it does not have sufficient information or confidence in a decision and refrain from issuing responses in such cases.

Unlike traditional "black box" AI systems, which generate results without revealing their internal logic, XAI works on the traceability, interpretability and accountability of these decisions. For example, if a neural network rejects a loan application, XAI techniques can highlight the specific factors that influenced the decision. Thus, while a traditional model would simply return a numerical rating of the credit file, an XAI system could also tell us something like "Payment history (23%), job stability (38%) and current level of indebtedness (32%) were the determining factors in the loan denial". This transparency is vital not only for regulatory compliance, but also for building user confidence and improving AI systems themselves.

Key techniques in XAI

The Catalogue of trusted AI tools and metrics from the OECD's Artificial Intelligence Policy Observatory (OECD.AI) collects and shares tools and metrics designed to help AI actors develop trusted systems that respect human rights and are fair, transparent, explainable, robust, safe and reliable. For example, two widely adopted methodologies in XAI are Local Interpretable Model-agnostic Explanations (LIME) and SHapley Additive exPlanations (SHAP).

• LIME approximates complex models with simpler, interpretable versions to explain individual predictions. It is a generally useful technique for quick interpretations, but not very stable in assigning the importance of variables from one example to another.
• SHAP quantifies the exact contribution of each input to a prediction using game theory principles. This is a more precise and mathematically sound technique, but much more computationally expensive.

For example, in a medical diagnostic system, both LIME and SHAP could help us interpret that a patient's age and blood pressure were the main factors that led to a diagnosis of high risk of infarction, although SHAP would give us the exact contribution of each variable to the decision.

One of the most important challenges in XAI is to find the balance between the predictive ability of a model and its explainability. Hybrid approaches are therefore often used, integrating a posteriori explanatory methods of decision making with complex models. For example, a bank could implement a deep learning system for fraud detection, but use SHAP values to audit its decisions and ensure that no discriminatory decisions are made.

Open data in the XAI

There are at least two scenarios in which value can be generated by combining open data with explainable artificial intelligence techniques:

• The first of these is the enrichment and validation of the explanations obtained with XAI techniques. Open data makes it possible to add layers of context to many technical explanations, which is also true for the explainability of AI models. For example, if an XAI system indicates that air pollution influenced an asthma diagnosis, linking this result to open air quality datasets from patients' areas of residence would allow validation of the correctness of the result.
• Improving the performance of AI models themselves is another area where open data brings value. For example, if an XAI system identifies that the density of urban green space significantly affects cardiovascular risk diagnoses, open urban planning data could be used to improve the accuracy of the algorithm.

It would be ideal if AI model training datasets could be shared as open data, so that it would be possible to verify model training and replicate the results. What is possible, however, is the open sharing of detailed metadata on such trainings as promoted by Google's Model Cards initiative, thus facilitating post-hoc explanations of the models' decisions. In this case it is a tool more oriented towards developers than towards the end-users of the algorithms.

In Spain, in a more citizen-driven initiative, but equally aimed at fostering transparency in the use of artificial intelligence algorithms, the Open Administration of Catalonia has started to publish comprehensible factsheets for each AI algorithm applied to digital administration services. Some are already available, such as the AOC Conversational Chatbots or the Video ID for Mobile idCat.

Real examples of open data and XAI

A recent paper published in Applied Sciences by Portuguese researchers exemplifies the synergy between XAI and open data in the field of real estate price prediction in smart cities. The research highlights how the integration of open datasets covering property characteristics, urban infrastructure and transport networks, with explainable artificial intelligence techniques such as SHAP analysis, unravels the key factors influencing property values. This approach aims to support the generation of urban planning policies that respond to the evolving needs and trends of the real estate market, promoting sustainable and equitable growth of cities.

Another study by researchers at INRIA (French Institute for Research in Digital Sciences and Technologies), also on real estate data, delves into the methods and challenges associated with interpretability in machine learning based on linked open data. The article discusses both intrinsic techniques, which integrate explainability into model design, and post hoc methods that examine and explain complex systems decisions to encourage the adoption of transparent, ethical and trustworthy AI systems.

As AI continues to evolve, ethical considerations and regulatory measures play an increasingly important role in creating a more transparent and trustworthy AI ecosystem. Explainable artificial intelligence and open data are interconnected in their aim to foster transparency, trust and accountability in AI-based decision-making. While XAI provides the tools to dissect AI decision-making, open data provides the raw material not only for training, but also for testing some XAI explanations and improving model performance. As AI continues to permeate every facet of our lives, fostering this synergy will contribute to building systems that are not only smarter, but also fairer.

Content prepared by Jose Luis Marín, Senior Consultant in Data, Strategy, Innovation & Digitalization. The contents and views reflected in this publication are the sole responsibility of the author.

There is no doubt that digital skills training is necessary today. Basic digital skills are essential to be able to interact in a society where technology already plays a cross-cutting role. In particular, it is important to know the basics of the technology for working with data.

In this context, public sector workers must also keep themselves constantly updated. Training in this area is key to optimising processes, ensuring information security and strengthening trust in institutions.

In this post, we identify digital skills related to open data aimed at both publishing and using open data. Not only did we identify the professional competencies that public employees working with open data must have and maintain, we also compiled a series of training resources that are available to them.

Professional competencies for working with data

A working group was set up in 2024 National Open Data Gathering with one objective: to identify the digital competencies required of public administration professionals working with open data. Beyond conclusions of this event of national relevance, the working group defined profiles and roles needed for data opening, gathering information on their roles and the skills and knowledge required. The main roles identified were:

• Role responsible: has technical responsibility for the promotion of open data policies and organises activities to define policies and data models. Some of the skills required are:
  
  • Leadership in promoting strategies to drive data openness.
  • Driving the data strategy to drive openness with purpose.
  • Understand the regulatory framework related to data in order to act within the law throughout the data lifecycle.
  • Encourage the use of tools and processes for data management.
  • Ability to generate synergies in order to reach a consensus on cross-cutting instructions for the entire organisation.
• Technical role of data entry technician (ICT profile): carries out implementation activities more closely linked to the management of systems, extraction processes, data cleansing, etc. EThis profile must have knowledge of, for example:
  
  • How to structure the dataset, the metadata vocabulary, data quality, strategy to follow...
  • Be able to analyse a dataset and identify debugging and cleaning processes quickly and intuitively.
  • Generate data visualisations, connecting databases of different formats and origins to obtain dynamic and interactive graphs, indicators and maps.
  • Master the functionalities of the platform, i.e. know how to apply technological solutions for open data management or know techniques and strategies to access, extract and integrate data from different platforms.
• Open data functional role (technician of a service): executes activities more related to the selection of data to be published, quality, promotion of open data, visualisation, data analytics, etc. For example:
  
  • Handling visualisation and dynamisation tools.
  • Knowing the data economy and knowing the information related to data in its full extent (generation by public administrations, open data, infomediaries, reuse of public information, Big Data, Data Driven, roles involved, etc.).
  • To know and apply the ethical and personal data protection aspects that apply to the opening of data.
• Data use by public workers: this profile carries out activities on the use of data for decision making, basic data analytics, among others. In order to do so, it must have these competences:
  
  • Navigation, search and filtering of data.
  • Data assessment.
  • Data storage and export
  • Data analysis and exploitation.

In addition, as part of this challenge to increase capacities for open data, a list of free trainings and guides on open data and data analyticswas developed. We compile some of them that are available online and in open format.

Figure 1. Table of own elaboration with training resources. Source: https://encuentrosdatosabiertos.es/wp-content/uploads/2024/05/Reto-2.pdf

INAP''s continuing professional development training offer

The Instituto Nacional de Administración Pública (INAP) has a Training Activities Programme for 2025, framed in the INAP Learning Strategy 2025-2028.. This training catalogue includes more than 180 activities organised in different learning programmes, which will take place throughout the year with the aim of strengthening the competences of public staff in key areas such as open data management and the use of related technologies.

INAP''s 2025 training programme offers a wide range of courses aimed at improving digital skills and open data literacy. Some of the highlighted trainings include:

• Fundamentals and tools of data analysis.
• Introduction to Oracle SQL.
• Open data and re-use of information.
• Data analysis and visualisation with Power BI.
• Blockchain: technical aspects.
• Advanced Python programming.

These courses, aimed at different profiles of public employees, from open data managers to information management technicians, allow to acquire knowledge on data extraction, processing and visualisation, as well as on strategies for the opening and reuse of open data in the Public Administration.  You can consult the full catalogue here..

Other training references

Some public administrations or entities offer training courses related to open data. For more information on its training offer, please see the catalogue with the programmed courses on offer.

• FEMP''s Network of Local Entities for Transparency and Citizen Participation: https://redtransparenciayparticipacion.es/.
• Government of Aragon: Aragon Open Data: https://opendata.aragon.es/informacion/eventos-de-datos-abiertos
• School of Public Administration of Catalonia (EAPC): https://eapc.gencat.cat/ca/inici/index.html#googtrans(ca|es
• Diputació de Barcelona: http://aplicacions.diba.cat/gestforma/public/cercador_baf_ens_locals
• Instituto Geográfico Nacional (IGN): https://cursos.cnig.es/

In short, training in digital skills, in general, and in open data, in particular, is a practice that we recommend at datos.gob.es. Do you need a specific training resource? Write to us in comments, we''ll read you!