Research Summary: A Reusable Semantic Web Framework for Evidence-Grounded Fundamental Rights Impact Assessments under the EU AI Act

A new Semantic Web-based framework has been developed to facilitate Fundamental Rights Impact Assessments (FRIAs) as mandated by Article 27 of the EU AI Act. This framework addresses the challenge of fragmented evidence by consolidating data from diverse sources into a SPARQL-queryable knowledge graph, specifically for high-risk AI systems in employment/worker management and access to essential public services. The framework, demonstrated with a 150-record corpus, aims to improve the credibility and efficiency of these legally required assessments.

Why it matters

The ability to conduct credible and efficient Fundamental Rights Impact Assessments is crucial for organisations deploying high-risk AI systems under the EU AI Act. This framework provides a structured approach to evidence consolidation, mitigating compliance risks and fostering responsible AI deployment across public sector domains.

Key insights

• The EU AI Act (Art. 27) mandates Fundamental Rights Impact Assessments (FRIAs) for high-risk AI systems prior to deployment.
• A significant challenge for credible FRIAs is the fragmentation of necessary evidence across various incident repositories, risk vocabularies, and legal texts.
• A reusable Semantic Web-based framework has been created to consolidate this fragmented evidence.
• The framework targets high-risk AI categories, specifically employment/worker management and access to essential public services.
• A curated 150-record corpus was annotated using keyword, LLM, and hybrid methods, resulting in a SPARQL-queryable knowledge graph of 1,351 RDF triples.
• Demonstration scenarios effectively surfaced relevant records (68.7% coverage) for FRIA purposes.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2609.39537