Trustworthy by Design: FAIR Data and Human-Centred AI Governance in CIRCMAN5.0
Technical blog post by Xgility Ltd (XGL)
CIRCMAN5.0 puts artificial intelligence into the hands of people who build, plan, and recycle photovoltaic products: AI-driven design tools, ML-assisted defect detection, and a human-in-the-loop recommendation engine for end-of-life recovery. An AI-aided framework is only as trustworthy as the data governance behind it. As leader of Task 1.4, Xgility is responsible for engineering that trust into CIRCMAN5.0 from day one.
FAIR Principles as an Engineering Requirement
The backbone of this work is the project’s Data Management Plan, built on the FAIR principles: data must be Findable, Accessible, Interoperable, and Reusable. In a project where four pilot factories in four countries feed data into shared digital twins, product passports, and a European data space, FAIR is what makes the architecture work. The plan defines how datasets are described and catalogued, which standards and ontologies make them interoperable across partners, and under what conditions industrial data can be shared, protected, or opened. First delivered at month 6 and updated at month 18, the plan evolves alongside the technical architecture, with a final update closing out the project.
Aligning with the EU AI Act
Manufacturing AI is entering a new regulatory era. CIRCMAN5.0 aligns its AI development with the High-Level Expert Group guidelines for Trustworthy AI and the requirements of the EU AI Act, whose obligations are progressively taking effect across Europe. In practice, this means the project’s AI components are assessed for human agency and oversight, transparency, robustness, and accountability, a natural fit for a project whose defining feature is keeping humans in the loop. The recommendation engine that suggests recovery strategies for end-of-life PV modules, for example, is explicitly designed so that experts review, adjust, and approve every recommendation before it is acted upon.
Governance as a Shared Culture
Trustworthy systems grow out of shared practice. Alongside the formal plan, Xgility runs data-sharing awareness workshops across the consortium, covering GDPR compliance, industrial data confidentiality, and the ethics of AI-assisted decision-making on the factory floor. The goal is simple: every partner handling pilot data should understand both the rules and the reasons behind them.
As circular manufacturing becomes data-driven, the projects that succeed will be the ones whose data can be trusted, shared, and reused. CIRCMAN5.0 is being built that way by design.