Trustworthy AI

Trustworthy Artificial Intelligence

Reliable, explainable, and responsible AI in critical domains

Explore branches Back to home

What is?

Trust in AI systems is not optional, it is fundamental. Trustworthy AI integrates explainability, privacy, equity, and robustness in the design, development, and deployment of intelligent systems. Our work aligns with European regulatory frameworks, especially the AI Act, ensuring that systems are safe, interpretable, and responsible in critical applications.

Pillars of Trust

Three fundamental axes that articulate our research in responsible AI

Explainability and Transparency

We develop methods so that AI systems are interpretable and auditable. Not just "black box results", but useful and verifiable explanations of decisions.

Equity and Privacy

We identify and mitigate biases, guarantee privacy in federated learning, and protect human autonomy against automated decisions.

Robustness and Security

Systems that work reliably in out-of-distribution contexts and critical domains. Governance aligned with European AI regulation.

Research Lines

Seven research lines aligned with European regulations and the needs of critical domains

Research Focus

Explainability (XAI)

Verifiable interpretability methods that go beyond visualizations: auditable and verifiable explanations.

Research Focus

Privacy & Federated Learning

Federated learning with differential privacy guarantees and distributed trust mechanisms.

Research Focus

Fairness & Bias

Bias detection and mitigation, fairness metrics, and preservation of human decision-making.

Research Focus

Robustness & OOD

Systems resilient to out-of-distribution data, adversaries and changes in operating environment.

Research Focus

Critical Applications

Safe AI deployment in healthcare, justice, finance and other high-risk sectors.

Research Focus

Governance & Regulation

Operational frameworks aligned with the European AI Act and best practices in responsible governance.