Today’s radar turns towards Africa and Asia in order to examine a less visible dispute than the race for ever larger models. The issue is not merely how to train more powerful systems, but how to decide which languages, memories and territories can enter a digital infrastructure without being reduced to examples, labels or administrable heritage.

The main signals come from a workshop on African language models at SOAS, from the celebration of World Kiswahili Language Day, from the rehabilitation of the National Museum of Sudan, from Mongolia’s first national geopark, from a conference on the future of Silk Roads heritage, and from the preparations for the next session of the World Heritage Committee in Busan. Read together, they point to a central problem for the anthropology of AI. The world does not fully fit the languages, categories and metrics through which institutions and technologies attempt to make it computable.

Africa and language models that do not fit the mould

On 8 and 9 July, SOAS and GSMA are holding the workshop Exploring African language models, devoted to bringing into dialogue AI research, African linguistics, translation, computer science and applied work with African languages. The call begins from a decisive observation: most large language model development has focused on a limited number of well-documented languages, especially European ones.

The anthropological interest of this scene lies not only in broadening linguistic coverage. The underlying issue is whether African languages must adapt to models designed from other linguistic ecologies, or whether their diversity forces the model itself to be reconsidered. The workshop call mentions questions such as the encoding of meaning, disinformation, benchmarks, multilingualism, code-switching, translanguaging, registers, style, community and endangered languages, ethics and research integrity.

AI does not merely “learn” languages. It also stabilises particular assumptions about what a language is, what counts as meaning, which forms of variation are treated as noise, and which communities are authorised to validate a translation, a classification or a response.

AIthropology Lab takeaway — Including African languages cannot mean merely adding corpora to a pre-existing architecture. It also requires asking who defines the correct language, which varieties are excluded, and who retains authority over meaning.

Kiswahili and digital linguistic justice

On 7 July, UNESCO placed Kiswahili at the centre of a discussion about the creator economy, digital public services and artificial intelligence. The organisation frames the issue as one of linguistic justice. Full participation in digital life should be possible in the language people know, use and pass on.

This case is especially relevant because it does not frame language as a residual heritage object, but as living infrastructure. Kiswahili appears in markets, classrooms, music, film, writing, code, platforms and public services. The issue is not limited to preserving a language. It also involves preventing digitalisation from turning inclusion into a promise conditional on the use of dominant languages.

For AIthropology Lab, this piece helps shift the usual debate on the digital divide. Connecting devices or translating interfaces is not enough. It is equally necessary to ask who defines the vocabularies of innovation, which institutions certify linguistic correctness, what data are deemed sufficient, and which forms of economic creativity remain outside the frame when a language is poorly represented in AI systems.

As broader context, UNESCO had already highlighted in 2025 the relationship between African languages and AI through experiences such as RobotsMali, which used ChatGPT and machine translation to produce more than 140 children’s books in Bambara. I do not treat this here as a new development of the day, but as an earlier signal of the same dispute: digital language revitalisation can reproduce hierarchies or open spaces for local reappropriation.

Sudan and the museum as memory under reconstruction

On 6 July, UNESCO published a note on the rehabilitation of the National Museum of Sudan, in collaboration with Sudanese institutions and Italian cooperation. The piece belongs to a context shaped by pandemic, political transition and armed conflict since April 2023, and presents cultural heritage as collective memory, identity, social cohesion and recovery.

This signal is not technological in the narrow sense, yet it helps widen the radar. Against the idea of heritage as a collection of objects, the Sudan case shows that a museum can also function as social infrastructure in the midst of crisis. Protecting heritage does not only mean preserving artefacts. It also means sustaining the conditions under which a community can narrate continuity, loss and reconstruction.

UNESCO itself documents concrete results. The project enabled a new exhibition proposal based on the documentation of more than 2,000 objects, interventions in security and conservation, specialised training, reports on illicit trafficking on online platforms and social media, training for more than 500 professionals, the recovery of 570 stolen objects, and the digital documentation of more than 10,000 items through new IT equipment.

Read alongside the block on African languages and AI, the point becomes clearer. Both the museum and the language model organise memory. Both select, classify, translate and circulate that which a society can recognise as its own. The museum simply makes the political fragility of that operation more visible.

AIthropology Lab takeaway — Digital preservation is not only archiving. In contexts of war, looting or displacement, it can also become an infrastructure of evidence, restitution and social continuity.

Mongolia and the territory that enters a global grammar

On 7 July, UNESCO published a note on the launch — held on 1 July in Khanbogd soum — of the Khanbogd – Shar Tsav Geopark, Mongolia’s first national geopark. The project brings together geological, palaeontological, natural and cultural heritage, with the participation of authorities, local communities, academia, international organisations and the private sector. It also prepares the ground for a possible future nomination to the UNESCO Global Geoparks network.

The news may seem distant from AI, yet it fits well within an anthropology of classification. A territory becomes a geopark when landscapes, fossils, nomadic memories, institutions, signage, tourism, science and local economy are assembled into a category that is globally recognisable.

The area extends across soums in the provinces of Umnugovi and Dornogovi, and includes fossil sites of major scientific importance as well as the Khanbogd massif, formed around 290 million years ago. The project is also supported by a sustainable tourism initiative led by UNESCO with funding from Rio Tinto Mongolia. That combination is precisely what makes the case interesting: conservation, science, tourism, private enterprise, local communities and international recognition appear as parts of the same territorial infrastructure.

Recognition cannot be assessed only in abstract terms of benefit or harm. It matters to observe what must be translated for a territory to become legible to UNESCO, to sustainable tourism, to science and to local development. It also matters to attend to what is left outside when territory becomes a heritage brand, a technical dossier and a promise of future value.

Silk Roads, living archive and digital technologies

Another UNESCO piece, published on 7 July, reports on the conference The Future of Silk Roads Heritage, held at Zhejiang University between 16 and 18 May. The meeting brought together 30 specialists from 10 countries and addressed shared heritage, multiple identities, conservation, historic cities in Central Asia, restitution of cultural property, living heritage, material culture and digital technologies.

The interest here lies in how a historical route is recomposed today as an archive, an academic network, a digital object and a grammar of international cooperation. The Silk Roads do not appear only as past history, but as a contemporary way of organising relations among research, cultural diplomacy, living heritage and interpretive technologies.

In this sense, the piece speaks to African languages, Sudan and Mongolia. All of them show that digitising is not simply a better way of conserving. It also means deciding which narrative organises memory, which scale is deemed legitimate, and which institutions can turn a multiplicity of practices into a recognisable object.

Busan as the prelude to a global politics of classification

The 48th session of the World Heritage Committee will be held in Busan from 20 to 29 July 2026, with the opening ceremony on 19 July. The committee will examine 30 new nominations, proposals relating to three already inscribed sites, and the state of conservation of 147 sites.

The list includes nominations from Asia and Africa that speak directly to this radar: Sarnath in India, Jingdezhen in China, Asuka and Fujiwara in Japan, Mangystau in Kazakhstan, the Fergana-Syrdarya corridor of the Silk Roads, the Xiongnu elite cemetery complexes in Mongolia, the historic medinas of the Comoros, the roças of São Tomé and Príncipe, Boma-Badingilo in South Sudan, and the renomination of Garamba in the Democratic Republic of the Congo.

Busan can function as a closing frame. World heritage does not only preserve. It also translates territories, memories, conflicts and ways of life into a global administrative language. That translation has material effects: it attracts protection, tourism, investment, prestige and surveillance. Yet it can also stabilise an authorised version of what a community, a landscape or a history ought to mean.

Closing

Today’s common thread is not technological novelty, but translation. African languages call for models that do not treat them as exceptions. Kiswahili shows that digital inclusion begins in everyday language. Sudan reminds us that memory is also protected as a condition of social reconstruction. Mongolia and the Silk Roads show how territories enter global categories of value. Busan anticipates the institutional scale of that operation.

No digital infrastructure works with bare data. It operates upon already hierarchised languages, already classified heritage, already translated territories, and communities that rarely have equal power to decide how they will be represented.

Sources