Built around interfaces
TRAILAB is not another specialist added to a project. It is built around the interfaces between technical evidence, legal duties, institutional governance, ethical questions and human agency.
Why TRAILAB
A research problem became too large for one discipline, one domain or one institution.
TRAILAB is a new independent institution built on mature research foundations. The trajectory behind it began with uncertainty, moved into high-stakes decisions, critical systems, causality, explainability, law, standards, ethics and human agency, and eventually required a durable home where those interfaces could be researched together.
The journey that made TRAILAB necessary
The journey that made TRAILAB necessary
-
2015-2019
Probabilistic doubt
The core technical question was how AI systems can represent uncertainty: domain limits, epistemic gaps and aleatoric variability, first motivated by financial and banking decisions where automation without doubt was dangerous.
-
2020-2022
From finance to critical systems
The same decision problem appeared in autonomous driving, aviation, industrial automation, fake-news detection and public projects: incomplete evidence, shifting operating conditions and human responsibility.
-
2021-2023
Causality and explainability
Uncertainty was not enough. The work expanded toward causality, counterfactuals and explainability: what caused a prediction, what would change under intervention, and whether explanations are faithful, stable and non-contradictory.
-
2023-2024
Law, AI Act and standards
Trustworthy AI entered regulated environments. AI Act obligations, ISO/IEC 42001, ISO/IEC 23894, CEN-CENELEC harmonised standards and sector assurance frameworks had to be translated into operational methods.
-
2024-2025
Ethics and philosophy
Technical and legal trust still left questions of legitimacy, fairness, care, authorship, responsibility, reputation and human purpose. Ethics and philosophy became part of the method, not an afterthought.
-
2026
Reflective AI and TRAILAB
As AI acquired infrastructural effects on judgment, Trustworthy AI had to be complemented by Reflective AI: a research objective for preserving and strengthening human capability. TRAILAB becomes the independent, multidisciplinary and multi-institutional space for that work.
Independent by design
Independence is valuable to partners and funders because it protects the credibility of methods, evidence and conclusions.
Applied by method
Impact happens when research is stress-tested in real systems, translated into standards and made reusable through projects, training and public-interest evidence.
Reflective by ambition
Reflective AI asks how reliance on AI changes the human capacities that make judgment, creativity and responsibility possible.
Our approach to impact
Our impact cycle turns research into capacity.
TRAILAB is designed so research, real-world validation and institutional transfer reinforce one another. Applied cases expose new research questions; methods are stress-tested in practice; policy and standards translate evidence into reusable capacity.
Research and regulatory anchors 9
-
2020 · Building Uncertainty Models on Top of Black-Box Predictive APIs
Uncertainty Quantification · Accepted paper
Brando, A., Torres, D., Rodríguez-Serrano, J. A., Vitrià, J.
-
2022 · Using Quantile Regression in Neural Networks for Contention Prediction in Multicore Processors
Uncertainty Quantification, Regulatory AI · Accepted paper
Brando, A., Serra, I., Mezzetti, E., Abella, J., Cazorla, F. J.
-
2023 · Main Sources of Variability and Non-Determinism in AD Software: Taxonomy and Prospects to Handle Them
Regulatory AI · Accepted paper
Alcon, M., Brando, A., Mezzetti, E., Abella, J., Cazorla, F. J.
-
2023 · SAFEXPLAIN: Safe and Explainable Critical Embedded Systems Based on AI
Explainability, Regulatory AI · Accepted paper
Abella Ferrer, J., Pérez, J., Englund, C., Zonooz, B., Giordana, G., et al. (incl. Brando, A.)
-
2024 · A Novel Application of SCMs to Time Series Counterfactual Estimation in the Pharmaceutical Industry
Causality · Workshop paper
Garriga, T., Sanz, G., de Cambra, E. S., Brando, A.
-
2025 · Managing Sources of Uncertainty in Utilizing AI in Development and Deployment of Safety-Critical Autonomous Systems
Uncertainty Quantification, Regulatory AI · Accepted paper
Lowe, R., Ulan, M., Bui, T. H., Adell, A., Labaien, J., Brando, A.
-
2026 · AI Sandboxes Must Be SME-Oriented Compliance Accelerators
Regulatory AI · Under review
Rodrigo-Albert, D., Eguíluz, J. A., Hidalgo, J., Vargas, V., Mir, M., Brando, A.
-
2026 · EU AI Act implementation timeline
Regulatory source · European Commission FAQ
The Commission frames the AI Act as a risk-based legal framework with staged application: prohibited practices and AI literacy from February 2025, GPAI governance from August 2025, and high-risk obligations now sequenced toward December 2027 and August 2028 under the simplification package.
-
2026 · Standardisation of the AI Act
Regulatory source · European Commission policy page
The Commission describes how harmonised standards translate AI Act requirements into technical language for risk management, data governance, logging, transparency, human oversight, accuracy, robustness, cybersecurity, quality management and conformity assessment.
Research → methods
Publications and open questions in uncertainty, causality, explainability, safety, regulation, ethics and Reflective AI.
Applied validation → evidence
Use cases in finance, banking, aviation, autonomous systems, automation, pharmaceutical research, health, education and public-interest infrastructures.
Systemic transfer → adoption
Regulation-aware methods, standards translation, policy capacity, SME adoption, public-sector guidance and rights-sensitive institutional learning.
TRAILAB double-cycle approach
Methods / know-how
uncertainty · causality · explainability · governance · reflective interaction
Research ↔ Application ↔ Impact
- ResearchPublications, doctoral work, benchmark problems, theoretical questions
- ApplicationBanking, finance, ADAS, aviation, pharma, health, automation, education
- ImpactAdoption, safer systems, better decisions, public-interest capacity
Research ↔ Systemic transfer ↔ Impact
- ResearchPublications, doctoral work, benchmark problems, theoretical questions
- Systemic transferAI Act context, ISO, CEN-CENELEC, public policy, civil society, rights
- ImpactAdoption, safer systems, better decisions, public-interest capacity
Why now - Europe
Europe is moving from AI rule-making toward large-scale adoption across strategic sectors while continuing to emphasise trustworthy, human-centric AI, skills and institutional capacity. TRAILAB works at the interface between capability, adoption, trust, rights and human agency.
- EU AI Act contextrisk-based duties, AI literacy, GPAI governance and high-risk system obligations
- Harmonised standardstechnical language for risk management, data governance, logging, transparency, human oversight, robustness and cybersecurity
- Institutional adoptiontranslation of capability into evidence, governance, skills and responsibility across sectors