CRAI – Center for Responsible AI

Brief and project description

The Center for Responsible AI is a national initiative that aims to position Portugal as a global leader in the technologies, principles and regulation of Responsible AI. Led by Unbabel, the consortium brings together a large ecosystem of startups, research institutes, a law firm and industry leaders, and is funded by the Portuguese Recovery and Resilience Plan (PRR). Its goal is to develop a new generation of AI products built on the technologies and principles of Responsible Artificial Intelligence, while creating a virtuous circle between startups, research centres and industry. The Faculty of Engineering of the University of Porto, through LIACC, is one of the research partners in this consortium.

The motivation behind Responsible AI is straightforward: as artificial intelligence becomes embedded in nearly every sector, so does the risk of unintended harm caused by opaque or poorly evaluated models. Making AI "responsible" means ensuring that its algorithms are as ethical, fair and transparent as possible, fostering public trust in the technology.

Within this broader effort, LIACC's contribution focuses on meta-learning and data generation and manipulation as a basis for building robust algorithm cards — standardised tools that document and evaluate models against Responsible AI principles such as fairness, transparency and accountability. By combining meta-learning with controlled data generation, the work aims to make model evaluation more rigorous, reproducible and resistant to the biases and blind spots that conventional benchmarks often miss.

The expected outcome is a set of methods and tools that help developers and organisations assess whether their models behave responsibly before deployment — turning abstract Responsible AI principles into concrete, measurable and actionable criteria.

Period: November 2022 – December 2025