WSIS Action Line C8 Cultural diversity and identity, linguistic diversity and local content

Valts Ernstreits argues that just as biological diversity keeps the planet healthy, linguistic and cultural diversity is essential to humanity's ability to innovate and move forward. The digital world has become a new reality permeating everyday life, and therefore a new domain has opened up in which linguistic and cultural diversity must be protected at least to the same degree as it was in the analogue world.

Ernstreits drew an explicit analogy between biological diversity and linguistic diversity, stating that diversity of languages and culture is relevant to humanity's ability to continue to innovate and move forward . He noted that the digital world is now present in buses, cars, and meeting rooms, meaning it is no longer a distant technology but a new domain requiring the same protections as analogue platforms .

Ernstreits highlights that a report on AI presented at the Global Forum on AI found that only 1,000 of the world's 7,000 languages have the foundations needed to enter AI domains. More alarmingly, the gap between those 1,000 languages and the remaining 6,000 is not narrowing, indicating a structural barrier that AI technology alone cannot solve.

He referenced a report on AI presented at the Global Forum on AI, which explicitly stated that out of 7,000 languages, 1,000 are prepared to enter AI domains, meaning 6,000 - the absolute majority - cannot . The report also noted that despite rapid advancements in AI, the gap between the 1,000 language-ready languages and the 6,000 excluded ones is not narrowing, pointing to a structural barrier rather than a technological one .

Ernstreits describes the Global Roadmap for Multilingualism in the Digital Era as one of the most important initiatives for breaking down the structural barriers preventing 6,000 languages from entering the digital domain. The roadmap, which emerged from the International Decade of Indigenous Languages, is structured around three dimensions: input (foundational requirements), output (technology integration), and process (community involvement).

He explained that the roadmap emerged from discussions during the International Decade of Indigenous Languages (2022-2032) and is structured around input, output, and process . He described input as covering basic requirements such as writing systems, standardisation, and the ability to produce digital data, which many of the 6,000 languages currently lack .

Ernstreits emphasises that the input phase of the Global Roadmap is extremely important because many of the world's 6,000 digitally disadvantaged languages lack even the most basic prerequisites for digital participation, such as writing systems and the ability to produce digital data. Without these foundational elements, no digital technology can be built for or used by these language communities.

He noted that speakers of major languages take for granted their ability to use and write in their language, but many of the 6,000 languages do not have writing systems or the ability to produce digital data, which is a precondition for any digital technology . He also mentioned that the input phase is diverse and includes very analogue aspects such as standardisation and tackling variation .

Ernstreits argues that even when language communities successfully develop their own language tools, they often cannot integrate these tools into mainstream software due to technical barriers. The solution he proposes is that technology should be multilingual by default, meaning it should be adaptable to any language from the outset.

He gave the personal example of working with the indigenous Livonian language, noting that even after building language technology, he cannot install a Livonian spell checker in word-processing software because the software simply does not allow it . He proposed that the key answer is for technology to be multilingual by default, so that it is adoptable to any language .

Ernstreits contends that language communities must be involved in the entire technology development process, from the initial idea through to deployment, not merely consulted at the end. This is justified both because they are rights and resource holders who need stewardship over their data and technologies, and because creating technology for endangered language communities is methodologically different and requires the communities' own knowledge.

He stated that language communities are rights holders and resource holders who need stewardship over data, solutions, and technologies developed for them . He also argued that creating technology for endangered communities is methodologically different, and communities possess the knowledge and understanding necessary for this process .

Ernstreits responds to the question about urbanisation and dialect loss by arguing that digital technologies can help by bringing together scattered communities that have been separated by urbanisation and population mobility. However, he stresses that for this to work, digital technologies must be capable of functioning for those specific communities.

He acknowledged that urbanisation is a challenge many communities are going through and that digital technologies can help by bringing scattered communities together . He also noted that there is insufficient understanding of how community-level decisions are made and how community governance is put into effect, identifying this as a further subject requiring exploration .

Ernstreits identifies a significant gap in current discussions about multilingualism and digital inclusion: while there is much talk about community-led governance and community involvement, there is not enough understanding of how community-level decisions are actually made and how community governance is operationalised in practice. He calls for this to be explored further.

He stated directly that there is a lot of talking about community-led governance and community involvement, but not enough understanding of how community-level decisions are made and how community governance is actually put into effect .

Canela highlights the enormous scale of the linguistic digital divide, noting that the vast majority of the world's languages lack even the most basic digital tools needed for meaningful participation in the digital world. This includes foundational elements such as keyboards, fonts, language datasets, and digital services.

He stated that more than 6,000 of the world's 7,000-plus living languages remain digitally disadvantaged , and that many still lack foundation tools, from keyboards and fonts to language datasets and digital services that make meaningful digital participation possible .

Canela argues that AI is not only a threat but also an extraordinary opportunity for language revitalisation. He outlines a range of specific ways in which AI can support endangered and low-resource languages, from documentation and translation to the creation of learning resources and digital preservation of oral traditions.

He listed specific AI-enabled opportunities including documenting endangered languages, improving communication, transcription and translation, creating language learning resources, expanding digital tools to support more languages and scripts, and preserving oral traditions through searchable digital archives .

Canela warns that AI also poses a significant risk of deepening existing linguistic inequalities if the languages and communities currently excluded from digital systems are not brought into the AI development process from the very beginning. He stresses that inclusion must happen at the outset, not as an afterthought.

He stated that AI risks reinforcing existing inequalities if digitally disadvantaged languages remain excluded from today's digital systems and the data that power tomorrow's AI, and especially if indigenous communities and other communities speaking these disadvantaged languages are not included in the conversation from the outset and not at the end of the process . He also noted that without investment in language documentation, digitalisation, community-led data, language infrastructure, and local capacity, AI cannot deliver inclusive outcomes .

Canela highlights that UNESCO's 2003 Recommendation on multilingualism in cyberspace was a pioneering normative instrument, approved by 194 UNESCO member states even before the WSIS process formally began in Geneva. After 23 years of implementation, there is now a strong and growing pick-up of interest in this topic, which should be leveraged.

He noted that the 2003 Recommendation was approved by 194 UNESCO member states even before WSIS started formally in Geneva, with negotiations having begun long before 2003 . He observed that after 23 years of implementing this recommendation, there is more and more interest in the topic, which is good news that should be leveraged .

Canela announces the launch of UNESCO's Quadriennial Report and Monitoring Mechanism for the 2003 Recommendation on multilingualism in cyberspace. He urges governments to participate in this exercise, emphasising that better evidence is essential for improving policies and enabling UNESCO to provide meaningful technical assistance.

He announced that UNESCO is launching the Quadriennial Report Mechanism and Monitoring Mechanism around the 2003 Recommendation and strongly encouraged governments to participate . He stressed that without data and evidence it is impossible to improve meaningful policies, and that the more governments report on what is happening, the better UNESCO can provide technical assistance .

Canela points out that both the WSIS Plus 20 review and the Global Digital Compact have reaffirmed that cultural diversity and linguistic inclusion are prerequisites — not afterthoughts — for meaningful connectivity. This makes UNESCO's normative work on multilingualism more relevant than ever in the current digital transformation landscape.

He stated that the WSIS Plus 20 review and the Global Digital Compact both reaffirmed that an inclusive digital future cannot be built on a monolingual foundation . He noted that UNESCO welcomes the strong emphasis placed by both processes on cultural diversity and linguistic inclusion as prerequisites, not afterthoughts, for meaningful connectivity .

Canela highlights the International Decade of Indigenous Languages as a successful model for community engagement, noting that indigenous communities are well connected and are themselves providing the key topics for the conversation on multilingualism. He suggests this offers a valuable learning experience for libraries, archives, and other institutions seeking to engage with language communities.

He pointed to the International Decade of Indigenous Languages as an interesting example of community engagement, noting that indigenous communities are really connected and are the ones providing the key topics for the conversation . He suggested there is a learning experience between these communities and libraries, archives, and similar institutions .

Xiaohong Hu emphasises that despite two decades of the WSIS process, global digital inequality and the linguistic digital divide continue to persist and are even expanding in the AI era. She notes that more than 2 billion people remain unconnected, underscoring the scale of the challenge.

She stated that compared to 20 years ago, global digital divide and global inequality continue to persist, and that even today more than 2 billion people are not connected . She also noted that the scale of the linguistic digital divide is outstanding and even expanding during the AI era .

Hu explains that the monitoring questionnaire for the 2003 Recommendation has been comprehensively updated to address new challenges in the digital and AI era. The updated questionnaire covers key areas such as language data representation, AI-related datasets, universal acceptance of domain names, and national frameworks for linguistic data governance.

She highlighted that the questionnaire includes key questions and indicators related to the representation of language data and content, AI-related datasets and language resources, the adoption of universal acceptance and international domain names, and national frameworks of linguistic data governance . She noted that all these dimensions closely contribute to the Action Line C8 .

Hu highlights UNESCO's role as host of the Secretariat of the UN International Decade of Indigenous Languages, which runs from 2022 to 2032. This initiative networks and empowers indigenous communities to preserve their languages, forming a key part of UNESCO's broader commitment to multilingualism.

She stated that UNESCO is hosting the Secretariat of the UN Initiative on International Decades of Indigenous Languages to network and empower indigenous communities and preserve their languages .

Hu notes that the Global Roadmap for Multilingualism in the Digital Era, with its 45 concrete stakeholder-driven actions, has received formal institutional recognition. The 35th Bureau of the UNESCO Information for All Programme has recommended that all member states and stakeholders implement the roadmap's actions across their different areas of work.

She stated that the roadmap, including its 45 concrete stakeholder-driven actions, has just been recognised by the 35th Bureau of UNESCO's Information for All Programme, which recommends all member states and stakeholders implement them in their different range of work .

Hu announces a new joint initiative between UNESCO, GovLab, and Microsoft called the New Commons Incubator for Indigenous Languages and Culture. This initiative directly enhances indigenous communities' capacity on data governance, and an open call for applications was at the time of the session.

She described the initiative as a global initiative on the New Commons Incubator for Indigenous Languages and Culture, launched in partnership with GovLab and Microsoft, which directly enhances community capacity on data governance . She noted that there was a deadline for applications before 14th August and called for engagement .

Elizabeth Oluoch argues that openness is a core technical principle of the internet, and that this openness is fundamentally incompatible with favouring a single language and script. She contends that language barriers artificially limit the internet's potential and stall its further growth, making multilingualism a benchmark of the internet's success.

She stated that the internet's technology is not simply a communications platform but aims to be a benefit to everyone through openness, and that we cannot maintain its core technical attribute of openness by favouring a single language and script . She argued that multilingualism on the internet expands the internet, while language barriers conversely stifle its further growth .

Oluoch outlines ICANN's substantial progress in implementing internationalised domain name standards over more than two decades. This work has resulted in over 150 IDN top-level domains representing 37 languages and 23 scripts, unlocking millions of domain names and allowing communities to connect and do business online in their native languages.

She noted that since 2003, ICANN has worked with the ICANN community to implement IDN standards, launching the IDN Country Code Fast Track Program in 2009, with 44 countries and territories now having allocated 63 country code top-level domains in local scripts . The 2012 new GTLD programme delegated 90 IDN generic top-level domains, resulting in over 150 IDN top-level domains representing 37 languages and 23 scripts, unlocking millions of domain names worldwide .

Oluoch argues that universal acceptance goes beyond being a mere technical requirement and is in fact an enabler of social and economic opportunity. By allowing multilingual internet users to access the internet and e-commerce in their native languages, universal acceptance opens up new markets and supports local communities.

She stated that universal acceptance is more than just a technical requirement and is an enabler of social and economic opportunity . She described how achieving universal acceptance readiness provides opportunities for global market expansion and a potential new client base by allowing multilingual internet users to access e-commerce and connect with local communities .

Oluoch uses the concrete example of international travellers to illustrate the real-world consequences of the lack of universal acceptance. Users with domain names and email addresses in non-Latin scripts encounter technical barriers when trying to access airport Wi-Fi, purchase tickets, or use other basic digital services, creating a digital divide in everyday situations.

She described how travellers using domain names and email addresses in their own scripts may encounter technical barriers, while those using standard English characters and traditional domain names like .com or .org navigate systems with ease . She noted that anyone using different scripts, IDNs, or newer top-level domains like .photography might get an error message and be unable to complete their transaction, and that this type of scenario happens every day for internet users around the world .

Oluoch highlights the recently published UNESCO-ICANN Policy Brief on Universal Acceptance as a practical and comprehensive resource for governments and the private sector. The brief provides a framework for accelerating digital inclusion and gives a sense of the 25-plus years of collaborative efforts by the technical community to enable multilingualism on the internet.

She stated that the universal acceptance policy brief developed by UNESCO in collaboration with ICANN offers a practical framework for governments and the private sector to accelerate digital inclusion . She noted that it gives a sense of the tremendous 25-plus years of collaborative efforts by the technical community, including the IETF, ICANN, and organisations such as UNESCO and Unicode to enable multilingualism on the internet .

Oluoch argues that achieving universal acceptance and digital inclusion requires coordinated action across multiple stakeholder groups. She outlines specific roles for governments, technology developers and service providers, civil society and academia, and international organisations, emphasising that each has a distinct contribution to make.

She stated that governments and regional organisations can embed universal acceptance and linguistic and script diversity into national digital strategies and public procurement, and lead by example in adopting universal acceptance-ready systems . She noted that technology developers and service providers should treat universal acceptance as a foundational design principle, not a niche feature , and that civil society, academia, and research organisations can raise awareness and advocate for multilingual technology in policy debates .

Monceau argues that AI has fundamentally changed the stakes of multilingualism by shifting from merely transmitting information to actively organising and generating knowledge. When languages are absent from AI training datasets, they do not merely become less visible online — they become less capable of contributing to the production of knowledge itself.

He stated that large language models no longer simply transmit information but increasingly organise knowledge, summarise it, recommend it, and generate it . He argued that when a vast majority of languages are absent from training datasets, they do not merely become less visible in cyberspace but become less capable of contributing to the production of knowledge itself .

Monceau contends that the rise of AI has elevated multilingualism from a cultural concern to a democratic, scientific, and geopolitical imperative. Because AI increasingly mediates how knowledge is produced, organised, and disseminated, ensuring linguistic diversity in AI systems is now a matter of fundamental importance beyond culture alone.

He stated that multilingualism is no longer only a cultural object but a social object, and has become a democratic, scientific, and geopolitical imperative . He argued that AI does much more than process language - it increasingly mediates the conditions under which knowledge is produced, organised, and disseminated .

Monceau warns that training AI systems predominantly on a small number of dominant languages risks not only reducing linguistic diversity but also homogenising human perspectives and ways of understanding the world. He argues that every language embodies a unique way of describing reality, and AI's dominance by a few languages threatens this epistemic richness.

He stated that every language embodies a particular way of describing the world and every culture carries unique categories of thought, memories, and forms of knowledge . He warned that if AI systems are trained predominantly on a handful of dominant languages, we risk reducing not only linguistic diversity but the diversity of human perspectives themselves .

Monceau argues that in the age of AI, creating multilingual cultural content is no longer sufficient — that content must also be discoverable. Recommendation algorithms, metadata, search engines, and generative AI increasingly determine what citizens encounter online, meaning that multilingual cultural expressions that are not discoverable gradually become invisible regardless of their quality.

He defined discoverability as the ability to find content online without necessarily searching for it, and argued that in the age of AI, creating cultural content is no longer sufficient - it must also be found . He noted that recommendation algorithms, metadata, search engines, and generative AI increasingly determine what citizens read, watch, and listen to, and that multilingual cultural expressions that are not discoverable gradually become invisible regardless of their intrinsic value .

Monceau highlights the OIF's role in ensuring that linguistic and cultural diversity is embedded in international digital governance frameworks. He notes that the OIF worked alongside member states and in close cooperation with UNESCO to ensure this principle is now reflected in the Global Digital Compact.

He stated that the OIF has worked actively to ensure that linguistic and cultural diversity became an explicit component of international digital governance, including during the negotiations of the Global Digital Compact, where this principle is now reflected . He also noted that the OIF has renewed its digital strategy to integrate artificial intelligence as a strategic priority .

Monceau describes the OIF's proactive steps in digital and AI governance, including the launch of a Francophonie Forum on Digital and AI Governance and the establishment of ongoing political dialogue among member states on AI governance. He frames this as reflecting the conviction that multilingualism is not about preserving the past but about preparing the future.

He stated that the OIF has launched a Francophonie Forum on Digital and AI Governance and established an ongoing political dialogue among member states on the governance of artificial intelligence . He expressed the conviction that multilingualism is not about preserving the past but about preparing the future .

Monceau argues that generative AI raises questions that did not exist when the UNESCO 2005 Convention was adopted, and that a new generation of international governance instruments is therefore needed. He specifically highlights a proposal to develop an additional protocol to the 2005 Convention addressing challenges such as training datasets, fair remuneration for creators, and transparency in recommendation algorithms.

He posed a series of questions that generative AI raises which the 2005 Convention did not address, including how to preserve linguistic diversity when AI models are trained on a limited number of languages, how to ensure fair remuneration for creators whose works contribute to AI training, and how to guarantee transparency regarding training datasets and recommendation algorithms . He highlighted the proposal from UNESCO's Reflection Group to develop an additional protocol to the 2005 Convention specifically addressing challenges created by the digital environment and AI .

An unnamed participant argues that raising awareness about linguistic diversity and digital inclusion must operate simultaneously at two levels: top-down, engaging with policy decision-makers and parliaments, and bottom-up, working directly with communities. They note that there are many available tools and reports, including the recently released scientific panel report on AI, that can support this awareness-raising effort.

The participant stated that raising awareness is key and must be both top-down, meaning working with policy decision-makers and parliaments, and bottom-up with communities, and that there are a lot of tools and networks available . They referenced the report released by the scientific panel on AI three or four days prior as a useful tool for raising awareness, particularly for showing the concentration of AI .

Stephen Wyber from IFLA raises the question of how to bring together local and community actors — including libraries — to take the decisions necessary to make multilingualism on the internet a practical reality. He notes that not all communities have access to established cultural institutions like the Alliance Française, and asks how community governance can be operationalised at the local level.

He noted that there is already a need in the plan on multilingualism to work with communities in a meaningful way to bring people together, and that not all communities are blessed with an institution like the Alliance Française . He asked specifically how local actors, community actors, and libraries can be brought together to take the decisions necessary to make multilingualism on the internet a reality .

David Castillo Parra frames the session by acknowledging the progress made under WSIS Action Line C8 in advancing multilingualism online, including internationalised domain names and multilingual content growth. However, he emphasises that significant gaps persist, particularly for indigenous peoples and speakers of low-resource languages who continue to face structural barriers. He also notes that the rapid rise of AI has deepened rather than closed these inequalities.

He stated that since the adoption of the WSIS Framework, Action Line 8 has provided the normative foundation for advancing multilingualism online, from internationalised domain names to the growth of multilingual content . He noted that indigenous peoples and speakers of low-resource languages continue to face structural barriers to meaningful digital participation, and that the rapid rise of AI has in many ways deepened these inequalities rather than closed them .

Castillo Parra contextualises the session within a broader set of recent international developments and normative outputs that together create a timely opportunity to advance multilingualism in the digital era. He highlights the convergence of multiple processes and documents that collectively reinforce the importance of the session's themes.

He stated that the session comes at a timely moment following the outcomes of the WSIS Plus 20, UNESCO's Global Roadmap of Multilingualism in the Digital Era, Mundia Code 2025, and the UNESCO-ICANN Policy Brief .

Castillo Parra, in his moderating role, reinforces the session's central thesis that digital inclusion is inseparable from multilingualism. He frames multilingualism not merely as a matter of content availability but as a prerequisite for meaningful participation in the digital society, including the ability to govern one's own data in relation to language models.

He introduced Guilherme Canela's framing that digital transformation cannot be truly inclusive if it is not multilingual , and the argument that multilingualism is ultimately about participation, ensuring people can access and shape the digital society in their own languages and scripts, and also govern their own data in their relationship with language models .

Castillo Parra makes a procedural point about ensuring that remote participants are not excluded from the discussion, reflecting the session's broader theme of inclusion. By explicitly inviting remote participants to submit questions via chat, he models the principle of inclusive participation that the session advocates for in the digital sphere.

He invited remote participants to submit their questions through the chat at any point, stating that he would bring them into the discussion during the Q&A segment .