Artificial intelligence (AI) is transforming how societies preserve and engage with intangible cultural heritage (ICH)—the living traditions, practices, and expressions that communities transmit across generations. From machine-learning restoration of artefacts to conversational systems that support oral traditions and language revitalisation, AI promises unprecedented visibility, access, and interactivity. Yet this promise sits alongside structural risks: gaps in data sovereignty, uneven participation in the AI lifecycle, and the commodification or misrepresentation of cultural expressions—especially those held by ethnic minority groups and Indigenous communities.
AI’s enhancement role: interactivity for living heritage
Recent initiatives illustrate AI’s constructive potential when culturally situated. In the Arab region, Arabic-centric large language models (LLMs) such as Fanar aim to treat Arabic and its dialects as primary rather than peripheral languages, preserving Islamic heritage and enabling culturally nuanced interaction at scale. In China, researchers use natural language processing and neural architectures to curate and recognise the multi-ethnic folk-song tradition Hufa’er, building a richer corpus for community learning and scholarship. Museums and educators increasingly deploy AI to personalise engagement; systems blending natural language processing, cognitive computing, and machine learning have delivered measurable gains in visitor participation, while extended reality platforms use AI services to weave emotionally resonant narratives from traditional music and stories. These advances matter because ICH is inherently dynamic—not a fixed “record of the past” but a living practice recreated through everyday use, pedagogy, and community transmission. When co-designed with communities, AI’s interactive affordances can serve that living quality.
Side effects: data, power, and cultural colonisation
At the same time, the AI lifecycle—from data collection to model training and deployment—is shaped by global inequalities. Online content is disproportionately produced in Western languages and reflects Western cultural values, while many Indigenous and minority communities face barriers to digital access and data control. Where internet connectivity is limited, communities have less capacity to participate in dataset design, audit model behaviour, or govern downstream uses. UNESCO’s Recommendation on the Ethics of Artificial Intelligence highlights inclusiveness, fairness, and cultural diversity, urging states to promote locally relevant systems and multilingual content. Yet ethical principles must be matched with institutional arrangements and power-sharing. Scholars caution against the dominance of governments in electronic ICH inventories and their exploitation, urging that civil society associations and organisations play fundamental roles in setting digital rules. Without such counterweights, AI risks enabling a form of digital cultural colonisation in which sacred motifs are used out of context, identities are flattened in training corpora, and communal meanings are abstracted into raw material for commercial systems.
Why copyright alone is not enough
Many communities seek intellectual property (IP) remedies to curb misappropriation. Yet copyright doctrine—built on individual authorship, originality, and fixation—rarely fits ICH’s collective, oral, and evolving nature. Consequently, most ICH materials are treated as part of the public domain, making them available for text and data mining (TDM) without consent or remuneration. By contrast, creative workers have launched campaigns and lawsuits contesting unconsented TDM of copyrighted works, calling for consent, credit, and compensation. This produces a stark asymmetry: ethnic minority groups and Indigenous communities lack comparable legal leverage to resist unrestricted computational use of their culture. Narrow TDM exceptions—for example, the United Kingdom’s non-commercial research carve-out—protect certain copyrighted works, while the European Union enables rightholders to opt out (Article 4(3)) via machine-readable reservations. Yet such mechanisms do not effectively reach ICH that is treated as non-copyrightable public domain.
AI and ICH as conceptual outsiders to copyright
Across Latin America and the Gulf States, TDM provisions are limited or silent, and data protection laws primarily target national security and privacy concerns rather than cultural self-determination. In practice, IP reform alone cannot resolve the mismatch: ICH’s communal authorship and AI’s distributed generativity both confound traditional copyright categories. Debates over AI-generated outputs—authorship, originality, and moral rights—reveal a persistent human-centric orientation in copyright across major jurisdictions. Courts and agencies emphasise human labour, skill, and a creator’s personal imprint, while generally rejecting non-human authorship. This human-centred logic mirrors the legal invisibility of ICH’s communal creativity, placing both AI and ICH in a conceptual outsider position within copyright law.
Towards responsible AI for cultural diversity
Recognising this parallel suggests that simply extending existing copyright categories will be insufficient. What is needed are multi-layered governance approaches combining community participation, data standards, and ethical AI with carefully targeted legal tools. A pragmatic agenda should focus on balance, not exclusion. Copyright has adapted to past technological shifts—from printing and broadcasting to digital reproduction—but AI’s speed and scale demand complementary measures. Responsible pathways include co-design with ICH-holding communities at the earliest stages of data collection and model evaluation; recognition of Indigenous data sovereignty norms that secure communal rights to collect, control, interpret, and benefit from data use; culturally aware data standards that require provenance, context metadata, and machine-readable reservations for cultural materials (extending the spirit of EU TDM opt-outs to non-copyright heritage assets); investment in non-English corpora and regional models with public funding tied to diversity benchmarks; and civil society stewardship to audit and govern digital ICH inventories so that neither state nor corporate actors monopolise cultural decision-making.
Conclusion
AI can enrich ICH by amplifying participation and making living traditions more accessible and engaging. It can also accelerate inequality when communities are excluded from data governance and their cultural expressions become extractive inputs. Because copyright’s individualist architecture struggles with both ICH and AI generativity, law alone cannot fix the problem. A hybrid governance approach—combining ethical AI, community participation, culturally aware data standards, and selective legal reform—offers a credible path forward. In practice, responsible AI for ICH means aligning technical progress with consent, credit, compensation, and the self-determination of the communities whose living cultures we seek to sustain.
Leave a Reply