AI in the Curator’s Loop: Designing Transparent and Trustworthy Metadata Displays under the EU AI Act

Museums are increasingly exploring how artificial intelligence could assist with largescale metadata enrichment, yet few have implemented such systems beyond small
research pilots. Concerns about hallucinations, bias, transparency, and the erosion of
curatorial authority have made many institutions cautious about operational use. This
paper introduces Project SPOT, a research-led framework and AI-assisted tool
developed to operate within the curator’s loop, ensuring that AI generated suggestions
support rather than supersede and replace curatorial knowledge and judgement. SPOT
identifies sub-objects within artefact images and produces candidate metadata that
are subsequently reviewed, amended, or rejected by curators, maintaining clear human
authorship throughout the process.
Forthcoming transparency obligations under the EU AI Act add urgency to these
discussions. The Act requires public disclosure when content has been generated or
influenced by AI, raising new design questions for museums that wish to display AIassisted information online. Using SPOT as a case study, this paper examines
strategies for integrating AI-generated, curator-verified data into artefact webpages
while remaining compliant with the Act. It considers both interface-level and metadatalevel approaches for signalling AI involvement and analyses how these might affect user
trust in institutions long regarded as sources of reliable knowledge.
By framing AI as a transparent, curator-controlled partner in the creation of cultural
data, the paper contributes to current debates on human–machine interaction, digital
provenance, and the museum as an epistemic actor in the digital public sphere.