Museums Rewrite the Ethical Rulebook as AI, Climate, and Colonial Legacies Move to Centre Stage
The International Council of Museums has adopted its first full code-of-ethics revision in two decades, with language on AI, climate, and restitution aimed at an institution in flux.

In the cavernous halls of the Louvre on 26 June 2026, the world's largest museum body quietly redrew the ethical perimeter of its profession. The International Council of Museums (ICOM), a UNESCO-affiliated network of roughly 50,000 institutions across 141 countries, formally adopted a revised Code of Ethics for Museums at the close of its 41st Ordinary General Assembly — the first comprehensive overhaul of the document since 2017, and arguably the most consequential since the original framework was written in 1986.
The new code arrives at a moment when museums are being asked, often uncomfortably, to be judges of their own inheritance. Generative artificial intelligence has begun populating collection databases and visitor apps. Insurance markets are repricing climate risk for heritage sites from Venice to Ulaanbaatar. And the politics of restitution — once a niche debate among curators — now routinely plays out in foreign ministries. ICOM's response, formalised in Paris, tries to set rules for a sector that has spent the better part of a decade improvising.
A code for a profession under pressure
The revisions touch three fault lines at once. The first is artificial intelligence. The code now obliges museums to disclose when AI tools are used to generate cataloguing metadata, fabricate reconstructions, or produce visitor-facing interpretation. It stops short of an outright ban — a position many institutions have lobbied for — but requires that human curatorial authority remain identifiable in the chain of authorship.
The second is climate. The new text commits museums to measuring and disclosing the carbon footprint of exhibitions, lending, and long-haul transport of works. It also frames museums as sites of climate literacy, not merely repositories of objects, and instructs institutions in vulnerable locations to develop disaster-response plans for their own collections.
The third is the post-colonial inheritance. The code strengthens language on provenance research, on partnership with source communities, and on the return of human remains and sacred objects — language that has been on the books for decades but was, in the working group's own framing, treated as aspirational by too many institutions. The Paris vote elevates it from principle to obligation.
ICOM president Emma Nardi, presenting the final draft in Paris, described the revision as a response to "the responsibilities that fall to museums in a world that has changed faster than our standards." The framing was deliberate: ICOM needed a code that could survive contact with technologies and political pressures the original drafters could not have imagined.
The counter-narrative: who pays, and who governs
Not everyone in the assembly hall was satisfied. Smaller institutions from the Global South warned, in working-group sessions that ran through the spring, that a code requiring carbon disclosure, AI audits, and provenance research would impose costs their budgets cannot absorb. The Association of African Museums, in a written submission circulated before the vote, argued that compliance obligations should be tiered — that a national museum in Lagos should not be benchmarked against the Smithsonian before basic digitisation has been completed.
ICOM's compromise language acknowledges the asymmetry. The code recommends, rather than mandates, certain disclosure practices for institutions under a staffing threshold, and it encourages partnership arrangements with better-resourced peers. Critics read that as a fig leaf; supporters read it as the only politically workable way to write a single global standard for an extraordinarily unequal membership.
The colonial-legacy provisions carry their own friction. European institutions have spent the last several years returning objects to Benin, Ethiopia, and, in fits and starts, across South Asia — but the pipeline of requests is now longer than the pipeline of transfers. The new code does not compel returns; it obliges museums to engage transparently with claims and to publish the criteria they use to decide them. That is a step short of what restitution campaigners want, and a step beyond what several national heritage agencies say their mandates permit.
Structural frame: the museum as regulator
What the Paris vote really ratifies is a shift in how museums see themselves. For most of the twentieth century, the profession's ethical gravity was internal — the care of objects, the integrity of scholarship, the relationship to the visitor. The 2026 code reaches outward: museums are now expected to make pronouncements about AI governance, climate disclosure, and the geopolitics of cultural heritage.
That shift has institutional consequences. Museums are, in practice, being asked to act as soft regulators in domains where national regulators have been slow to move. The European Union's AI Act, which began phased application in 2025, exempts most cultural-heritage uses from its highest-risk classification — a gap the ICOM code now fills, in part, with self-imposed disclosure duties. On climate, museums were among the first major cultural institutions to commit to emissions reporting frameworks that pre-date binding national rules in several jurisdictions.
There is an irony in this. A profession built on stewardship of the past is now being drafted into governance of the present. The risk is that museums, under-resourced and politically exposed, end up carrying obligations their public funders have not agreed to fund. The countervailing argument is that no other institution sits at the same intersection of public trust, technical authority, and global reach.
Stakes and forward view
The first test of the new code will be mundane: whether individual institutions actually update their acquisition, lending, and digital-publication policies to match. ICOM has no enforcement mechanism; its authority runs through national committees and through the reputational weight of the network. For the code to bite, funder behaviour — government ministries, private foundations, corporate sponsors — has to reward compliance and penalise its absence.
The harder tests are political. If a major museum refuses an AI-disclosure request from its national press, the framework will be tested in public. If a provenance claim from a source country is rejected without published reasoning, the new transparency language will be cited in the response. Over the next eighteen months, watch how ICOM's ethics committee handles its first formal complaints — the cases it accepts, and the ones it declines.
The 2026 code will not end the debates over restitution, AI authorship, or climate liability in cultural work. What it does is move those debates inside a written standard, signed by the profession's largest representative body, that cannot be ignored by a board of trustees looking for cover. For an institution type accustomed to operating above the political fray, that is a quieter but more durable form of accountability than any single piece of returned art or any single algorithmic disclosure.
This piece sits on the culture desk. Where wire coverage has tended to treat the ICOM revision as a procedural update, Monexus reads it as a governance inflection — museums being asked, by themselves, to write rules their national regulators have not.
Wire provenance
This editorial synthesis draws on the following public wire/social posts:
- https://en.wikipedia.org/wiki/International_Council_of_Museums
- https://en.wikipedia.org/wiki/Code_of_Ethics_for_Museums_(ICOM)
- https://en.wikipedia.org/wiki/Emma_Nardi