AI Resilience Report forCurators

Right now, AI in the museum world is mostly augmenting curators rather than replacing them. Behind the scenes, AI is speeding up the most repetitive parts of the job — cataloging, metadata, and database management. The U.S. Institute of Museum and Library Services just put more than $4 million behind eight projects to build AI literacy and integrate AI tools into museums and libraries nationwide ^[1], a federal push that signals AI is becoming a normal part of collections work.

AI-generated art is also stepping into the spotlight: Dataland, billed as the world's first museum of AI art, will open its inaugural "Machine Dreams: Rainforest" exhibition in downtown Los Angeles on June 20, 2026 ^[2], using a model trained on millions of nature images. But the heart of curating — choosing themes, interpreting objects, writing grants, building community — is still very human. As one American Alliance of Museums essay puts it, object-based work calls for "embodied experience and contextual nuance" ^[3] that AI queries cannot replicate.

Adoption is moving forward, but carefully. A new Audiences Agency report shared by the Museums Association brought together 32 leaders and practitioners from 16 UK cultural organisations to experiment, learn and strategise around the use of artificial intelligence ^[4] — and urged the sector to keep adoption "people-centred, not led by AI." Cost is dropping because tools are widely commercial and grants are growing, but several brakes apply: the same report flags the environmental cost, baked-in bias within large language models, and the risks to the creative arts posed by generative AI ^[4] as serious unresolved issues. ICOM UK is currently running an international study exploring how AI is used in everyday practice, how it is perceived, and how it shapes communication, trust, and institutional voice ^[5], which shows how seriously the field is taking ethics.

And Brookings cautions that occupations with high "AI exposure" don't always show high actual usage ^[6] — meaning the jump from "AI could do this" to "AI is doing this" is slower than headlines suggest. For curators, that's actually hopeful news: your judgment, storytelling, and trust-building skills are exactly the parts machines struggle with most.