ILO urges skills investment as AI reshapes ASEAN workforce

The International Labour Organization (ILO) has published a report examining how generative AI (GenAI) is reshaping labour markets across the Association of Southeast Asian Nations (ASEAN).

The study estimates that nearly 80 million workers, representing 22.9% of total employment in the region, work in occupations with more than minimal potential exposure to GenAI. However, the ILO stresses that there is currently no evidence of large-scale job displacement.

Only 3.3% of ASEAN workers, around 11.7 million people, are employed in occupations with the highest level of GenAI exposure, while roughly two-thirds of employment remains in occupations with no identified exposure.

Employment in highly exposed occupations has continued to grow, suggesting that AI is transforming work rather than replacing jobs at scale. The report also notes that adoption remains concentrated in technology-intensive sectors and is still relatively limited in many administrative occupations despite their high exposure.

The report identifies significant differences across ASEAN economies. Singapore has the highest share of workers with more than minimal GenAI exposure at 42.2%, followed by the Philippines (28.1%), Indonesia (21.7%), Vietnam (20.8%), and Thailand (20.6%).

The ILO also highlights a notable gender gap, with women more than twice as likely as men to work in highly exposed occupations because they are more heavily represented in clerical, administrative and professional roles. By contrast, exposure levels are broadly similar across younger and older working-age groups.

To maximise the benefits of AI while limiting potential risks, the ILO calls for human-centred AI governance, expanded upskilling and reskilling programmes, stronger support for micro, small and medium-sized enterprises (MSMEs), and closer regional cooperation on skills development.

The report argues that future labour market outcomes will depend less on AI exposure itself than on policy choices that strengthen the preparedness and resilience of workers, businesses and institutions.

Why does it matter?

The report challenges the assumption that generative AI will rapidly eliminate large numbers of jobs across Southeast Asia. Instead, it suggests AI is more likely to reshape existing occupations, with the scale of change depending on how quickly workers, businesses and governments adapt.

The findings also highlight that AI adoption is ultimately a policy challenge as much as a technological one. Investments in skills, workforce transitions and responsible AI governance will play a decisive role in determining whether AI improves productivity and job quality or widens existing inequalities across the region.

Would you like to learn more about AI, tech and digital diplomacyIf so, ask our Diplo chatbot!

UN explores how AI can scale human rights implementation

Digital tools and AI can help governments turn thousands of human rights recommendations into concrete action, but only if technology remains firmly guided by human expertise and institutional cooperation, speakers concluded during a WSIS Forum 2026 session on scaling digital tools for human rights monitoring.

The discussion brought together representatives from Costa Rica, the Office of the UN High Commissioner for Human Rights (OHCHR), academia, and civil society to examine how digital platforms, AI-assisted analysis, and improved data management can enhance the implementation of recommendations issued by UN human rights mechanisms.

Costa Rica shares experience with recommendation tracking

Opening the discussion, Domenico Zipoli, Head of Programmes at the Geneva Human Rights Hub, noted that governments receive thousands of recommendations every year from treaty bodies, the Universal Periodic Review (UPR), special procedures, and regional mechanisms, making implementation increasingly difficult without digital support.

Costa Rica’s Roberto Cespedes, Chargé d’Affaires at the country’s mission to the UN in Geneva, explained how the National Recommendations Tracking Database (NRTD) has transformed the country’s follow-up process.

Costa Rica established its National Mechanism for Implementation, Reporting and Follow-up (NMIRF) in 2011, bringing together ministries, parliament, the judiciary, and the national human rights institution. However, for years, the mechanism lacked an effective technological platform capable of managing recommendations from multiple international processes.

‘The database has significantly improved visibility of recommendations across institutions,’ Cespedes said.

He highlighted the tool’s ability to cluster recommendations by topic, enabling ministries to identify shared responsibilities and collaborate more effectively. Rather than working in isolation, institutions increasingly recognise the need for coordinated implementation.

Costa Rica is also working to expand access beyond government. Cespedes said civil society organisations are expected to gain direct access to the platform, allowing them to monitor implementation, provide feedback, and strengthen transparency.

OHCHR: AI can assist, but humans remain indispensable

Presenting the UN perspective, Marie Eve Boyer, Human Rights Officer at OHCHR, explained that the NRTD was developed to address the fragmentation of international human rights recommendations.

Built on the Universal Human Rights Index, the platform enables governments to consolidate recommendations, assign responsibilities across ministries, monitor progress, and prepare reports more efficiently.

Boyer noted that 20 countries are already using the NRTD, while another 40 are waiting for deployment.

She argued that AI has significant potential to support implementation by identifying relevant information, clustering recommendations, highlighting data gaps, and scaling reporting processes. However, she stressed that technology cannot replace human judgement.

‘AI can help process information, but it cannot understand the reality experienced by communities,’ she said, adding that contextual expertise remains essential when assessing whether recommendations have genuinely been implemented.

She also warned against viewing digital tools as substitutes for strong institutions, arguing that successful implementation depends on sustained human engagement alongside technological innovation.

Generative AI opens new possibilities for legal experts

Offering an academic perspective, Lukasz Szoszkiewicz, Assistant Professor at Adam Mickiewicz University in Poznań, demonstrated several prototype tools built using natural language processing and generative AI.

His projects include searchable databases of UN treaty body jurisprudence, analytical dashboards for the Universal Human Rights Index, and paragraph-level search tools for European Court of Human Rights decisions.

Szoszkiewicz argued that generative AI is fundamentally changing software development by enabling lawyers, researchers, and other domain experts to build specialised digital tools themselves rather than relying solely on IT teams.

‘Domain experts now have the possibility to develop tools that match exactly what they need,’ he explained.

He also addressed concerns about AI hallucinations, recommending that large language models be used primarily to generate deterministic software code rather than directly analysing sensitive datasets. This approach, he said, produces more reliable and verifiable results while reducing the likelihood of inaccurate outputs.

Better data still needed to measure real-world outcomes

Audience interventions highlighted persistent challenges surrounding data availability and measuring whether human rights recommendations actually improve people’s lives.

Representatives from civil society organisations working on torture prevention and disability rights pointed to the difficulty of obtaining reliable outcome data, particularly in countries where governments do not systematically publish relevant information.

Responding to these concerns, Boyer said OHCHR is exploring minimum datasets that could help governments monitor implementation more consistently while aligning human rights indicators with the Sustainable Development Goals.

Cespedes added that AI could eventually help governments identify positive actions that officials may not even realise correspond to international recommendations, making implementation more visible and easier to document.

Throughout the session, speakers agreed that AI and digital platforms should be viewed as tools to strengthen human rights implementation rather than replace human oversight. They concluded that meaningful progress will depend on better data, stronger institutional cooperation, and continued collaboration between governments, international organisations, academia, and civil society.

Track all key moments from the WSIS Forum 2026 on our dedicated WSIS page.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot

Universal acceptance key to multilingual internet, UNESCO and ICANN say

The internet cannot be truly inclusive until every valid domain name and email address works regardless of language or script, speakers said during a WSIS Forum 2026 session marking the launch of UNESCO and ICANN’s joint policy brief on universal acceptance (UA). The discussion brought together representatives from UNESCO, ICANN, and the Organisation Internationale de la Francophonie (OIF), who argued that multilingual internet infrastructure has become increasingly important not only for digital inclusion but also for the future development of AI.

While internationalised domain names (IDNs) and multilingual email addresses have long been supported by global technical standards, implementation remains far from complete. Speakers stressed that closing this gap now requires stronger political commitment, procurement policies, and coordinated action from governments, industry, academia, and civil society.

Technical standards exist, but implementation lags

Opening the session, UNESCO’s Guilherme Canela argued that multilingualism is about much more than preserving languages, it is about enabling full participation in digital society.

‘A multilingual internet expands access to information, education, services, innovation, and opportunity,’ he said, noting that many websites and online services still fail to recognise or process domain names and email addresses written in local languages and scripts.

ICANN’s Theresa Swinehart echoed that message, describing universal acceptance as a practical challenge that can be solved rather than a technological limitation. She highlighted ICANN’s work on multilingual internet infrastructure, including the delegation of more than 150 internationalised top-level domains covering 37 languages and 23 scripts, alongside efforts to improve compatibility across software platforms and open-source projects.

Dr Sarmad Hussain, Senior Director for IDN and UA Programmes at ICANN, illustrated the scale of the remaining challenge with new data. A survey of around 1,000 websites across 20 countries found that, on average, only 12% accepted email addresses written in local languages. Meanwhile, fewer than 30% of the world’s email servers currently support internationalised email addresses.

‘The technology and standards already exist,’ Hussain explained. ‘The problem is that many applications and websites have simply not been updated to support them.’

UNESCO–ICANN policy brief offers roadmap

A central focus of the session was the launch of the joint UNESCO–ICANN policy brief Advancing Universal Acceptance of All Domain Names and Email Addresses for Multilingual Internet.

Presenting the document, UNESCO’s Xianhong Hu argued that achieving universal acceptance requires much more than technical upgrades. The policy brief identifies gaps in awareness, policy, capacity development, and implementation, while providing tailored recommendations for governments, international organisations, civil society, academia, technical communities, and language communities.

Hu also stressed that multilingual internet infrastructure has become increasingly important in the AI era.

‘Without universal acceptance, AI systems learn from a narrower and less representative digital environment,’ she said, warning that languages excluded from today’s internet infrastructure risk remaining underrepresented in tomorrow’s AI models.

The brief also introduces measurement frameworks that governments and organisations can use to monitor progress. UNESCO plans to integrate UA into the sixth monitoring cycle of its Recommendation on the Promotion and Use of Multilingualism, with national reports due in 2027.

Multilingual internet increasingly linked to AI governance

Several speakers argued that universal acceptance should no longer be viewed as a niche technical issue but as part of broader AI governance discussions.

Henri Monceau of the Organisation Internationale de la Francophonie said multilingual internet infrastructure directly influences which languages become visible in AI training datasets.

‘The richer and more linguistically diverse the internet becomes, the richer and more representative AI systems can become,’ he said, warning that languages marginalised online today may also be marginalised in future AI applications.

Speakers therefore called for governments to include universal acceptance in national digital transformation and AI strategies, alongside procurement requirements that encourage software developers to build multilingual support from the outset.

ICANN also highlighted its growing capacity-building efforts, including more than 200 UA Day events organised across 86 countries in 42 languages, reaching over 29,000 participants. The organisation is also working with universities in Bahrain, Mexico, India, Zimbabwe, Bolivia, and other countries to integrate universal acceptance into computer science curricula so future software developers build multilingual support by default.

End users and local communities remain central

Audience questions focused on ensuring that end users, not only technical communities, remain at the centre of universal acceptance efforts, as well as whether growing interest in AI sovereignty could conflict with broader multilingual accessibility.

Responding to these concerns, Canela argued that locally developed AI models and universal acceptance should reinforce rather than compete with one another. Smaller language models designed for specific communities, he said, can strengthen both linguistic diversity and digital inclusion while requiring fewer computing resources.

Panellists agreed that the success of universal acceptance ultimately depends on collaboration across governments, industry, academia, civil society, and local language communities. They concluded that making every valid domain name and email address work equally across the internet is no longer simply a technical objective, but a prerequisite for a multilingual, inclusive, and AI-ready digital future.

Track all key moments from the WSIS Forum 2026 on our dedicated WSIS page.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot

African leaders push for homegrown AI and value creation at WSIS Forum

African experts and industry leaders used the WSIS Forum 2026 to argue that the continent must move beyond digital inclusion towards digital sovereignty, calling for greater investment in industrial capacity, locally developed AI, and value creation from Africa’s own resources and data rather than continued dependence on foreign technologies.

The session, ‘From Digital Inclusion to Digital Sovereignty: Building Capacity, Infrastructure, and Governance for Sustainable Digital Transformation,’ explored how Africa can become not only a user of AI and Industry 4.0 technologies, but also a producer of digital value. Moderated by Adelina Zeqiri of the University of Côte d’Azur, the discussion featured Professor Sama Mbang, Jean Bosco Byiringoro, and Professor Adel Ben Youssef, all founding members of the Alliance for Industry 4.0 and Smart Manufacturing in Africa (ASMA).

Industrialisation remains the foundation of development

Opening the discussion, Professor Sama Mbang argued that Africa risks falling further behind unless it accelerates industrialisation alongside digital transformation.

Drawing on his experience implementing Industry 4.0 solutions in manufacturing, Mbang stressed that industrial development remains the common denominator among prosperous economies.

‘There is no developed country that is not industrialised,’ he argued, adding that industrialisation creates the skills, technology, and productive capacity needed for long-term prosperity.

Mbang introduced ASMA as a platform designed to connect governments, industry, academia, and technical experts around practical projects in smart manufacturing, health, mining, automotive production, agriculture, and digital technologies.

He also highlighted the continent’s long-standing imbalance in global value chains. Although Africa possesses around 68% of the world’s critical minerals, it captures less than 1% of the value added from their processing. Similar disparities exist in pharmaceuticals, where Africa exports raw materials while importing most finished medicines.

According to Mbang, AI should support industrialisation, not replace it.

‘Sometimes talking about AI shifts attention away from the real challenge,’ he observed. ‘Africa first needs the capability to manufacture and transform locally.’

Building African AI for African realities

The discussion repeatedly returned to the distinction between adopting AI and developing AI that reflects African contexts.

Jean Bosco Byiringoro, professor of mechatronics and founder of ASMA, argued that importing models developed elsewhere will not solve Africa’s development challenges because they are built for different industrial environments.

‘What we need is not to import the model,’ he said. ‘We need to build our own model in the African context.’

Byiringoro argued that human capital is the continent’s greatest priority. Rather than focusing solely on software, African countries need engineers, technicians, manufacturers, and researchers capable of building AI systems rooted in local industries and value chains.

He illustrated this through agricultural projects that use digital representations of industrial equipment to help farmers understand production processes and develop new business opportunities. His organisation has already helped more than 2,000 people move into industrial employment through such initiatives.

Africa’s resources create new opportunities

Professor Adel Ben Youssef challenged participants to avoid viewing Africa as a single market, reminding the audience that the continent comprises 54 countries with diverse economic realities.

He nevertheless identified several shared competitive advantages.

Africa’s rapidly growing population, abundant renewable energy resources, and what he described as a ‘last mover advantage’ could allow countries to leapfrog older industrial models and build more sustainable digital infrastructure.

Rather than remaining dependent on foreign data centres, Ben Youssef argued that Africa could become a global location for digital infrastructure powered by renewable energy.

‘The real obstacle is not energy,’ he said. ‘It is political stability.’

He also warned that Africa’s creative industries face a growing threat as cultural content, artistic works, and local knowledge are increasingly used to train AI models without consent or compensation.

‘Most African creative content is being scraped to train AI models,’ he noted, arguing that this represents both an economic and cultural sovereignty challenge.

Human capital before regulation

Audience questions turned to data governance, with participants asking whether Africa should pursue GDPR-style regulation to protect its growing digital economy.

The panellists urged caution.

Ben Youssef argued that simply copying Europe’s General Data Protection Regulation would ignore Africa’s very different economic realities, particularly the importance of informal economies.

Instead, he called for flexible, sector-specific governance frameworks adapted to local contexts and accompanied by fair mechanisms for sharing the economic value generated from African data.

Mbang went further, arguing that the continent’s immediate priority should be creating value rather than replicating regulatory frameworks developed elsewhere.

‘Our fight today is not GDPR,’ he said. ‘Our fight is creating value locally instead of exporting raw materials and importing finished products.’

Byiringoro agreed, insisting that strong regulation can only emerge once countries have developed the human capital and industrial capabilities worth protecting.

Digital sovereignty through collaboration

The discussion concluded with an invitation to governments, universities, businesses, and international organisations to participate in ASMA’s growing network, including its inaugural conference in Dakar later this year.

While the session focused on Africa, speakers stressed that digital sovereignty should not be viewed as economic isolation or geopolitical competition.

Instead, they argued that enabling Africa to capture more value from its own resources, industries, and knowledge would strengthen global prosperity rather than diminish it.

Across the discussion, a consistent message emerged: AI alone will not transform Africa unless it is accompanied by investment in manufacturing, skills, infrastructure, and local innovation. For the panellists, digital sovereignty begins not with owning algorithms, but with building the industrial and human foundations that allow countries to shape their own digital future.

Track all key moments from the WSIS Forum 2026 on our dedicated WSIS page.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot

WSIS Forum 2026 opens with calls to turn digital commitments into action

The WSIS Forum 2026 opened in Geneva with a high-level appeal for stronger international cooperation to ensure that AI and digital transformation benefit everyone, not just the countries leading the technology race. Leaders from governments, the UN, academia, and civil society argued that the next phase of the World Summit on the Information Society (WSIS) must focus on implementing long-standing commitments on connectivity, digital inclusion, and AI governance rather than creating new principles.

Moderated by ITU Secretary-General Doreen Bogdan-Martin, the opening plenary brought together UN General Assembly President Annalena Baerbock, Estonian President Alar Karis, Kazakhstan’s Deputy Prime Minister Zhaslan Madiyev, Nobel Peace Prize laureate Kailash Satyarthi, and AI Academy Asia CEO Bolor-Erdene Battsengel to discuss how leadership can shape a more inclusive digital future.

Multilateral cooperation remains essential

Opening the discussion, Annalena Baerbock warned that multilateralism is under increasing pressure, making the recently adopted WSIS+20 consensus resolution an important demonstration that countries can still work together on digital issues.

She argued that AI governance cannot be separated from broader development challenges, stressing that discussions about responsible AI have little meaning where people still lack reliable internet access or electricity.

‘We can have the best AI governance systems in the world, but they will not matter if millions remain disconnected,’ she suggested, pointing to Tanzania’s digital health initiatives, which have connected almost two million people with healthcare services through WSIS-supported projects.

Baerbock also called for stronger cross-regional partnerships and warned that digital inequality increasingly intersects with broader challenges relating to development, peace, and human rights.

Estonia outlines principles for trusted digital governance

President Alar Karis shared Estonia’s experience as one of the world’s most digitally advanced governments, presenting five principles for building trusted digital societies.

According to Karis, digital infrastructure should remain open, secure, and interoperable, while governments should embrace inclusive multistakeholder governance involving the private sector, civil society, academia, and technical communities. He also stressed that human rights must be protected online just as they are offline, digital development should include skills and literacy alongside connectivity, and global initiatives such as WSIS and the Global Digital Compact should reinforce rather than duplicate one another.

Karis also highlighted Estonia’s investments in AI education, noting that all upper secondary school teachers and students are now being introduced to AI tools and literacy as part of a nationwide programme.

Kazakhstan showcases rapid digital transformation

Kazakhstan’s Deputy Prime Minister Zhaslan Madiyev outlined his country’s digital transformation strategy, describing digital infrastructure as the foundation for economic growth.

More than 90% of Kazakhstan’s public services are now available online, he said, supported by a digital ecosystem that includes over 2,000 technology companies and dedicated digital leadership across government ministries.

Madiyev also highlighted recent legislative reforms, including a constitutional amendment protecting digital rights and personal data, alongside plans to build one gigawatt of AI computing capacity within the next three to five years.

He argued that AI should increasingly be viewed as basic infrastructure, comparable to electricity, water, and internet connectivity, rather than simply another emerging technology.

Compassion must become part of AI

The session’s strongest moral appeal came from Nobel Peace Prize laureate Kailash Satyarthi, who challenged participants to think beyond technical capabilities and focus instead on whom AI ultimately serves.

Satyarthi argued that technology is never neutral because it reflects the values of those who create it. He urged developers to embed compassion, justice, and human dignity directly into AI systems, particularly for the benefit of vulnerable children.

One of his most striking proposals was that AI engineers should spend time working with children living in poverty, conflict zones, or remote communities before designing new systems.

‘When they return to their laboratories,’ he suggested, ‘they will write different code.’

His proposal received immediate support from Bolor-Erdene Battsengel, who said she would gladly encourage her own engineers to participate.

AI skills become the new economic infrastructure

Battsengel argued that digital inclusion today depends as much on skills as on connectivity.

Drawing on AI Academy Asia’s work across Mongolia and Central Asia, she described how training around 1,000 teachers enabled those educators to introduce AI tools to approximately 50,000 children living in remote communities.

Rather than treating AI education as a standalone technical programme, she described it as an investment in future economic competitiveness.

‘We no longer simply train people to use AI,’ she explained. ‘We build economic opportunity.’

Kazakhstan similarly reported training around one million people in AI-related skills during the past year and announced plans to launch a dedicated AI University later in 2026.

From dialogue to delivery

Closing speakers from UNESCO, UNCTAD, and UNDP argued that the international community should now shift from discussing digital inclusion to implementing it.

UNESCO stressed that people, not technology, remain at the centre of the WSIS vision, while UNDP highlighted ongoing work supporting national AI strategies and public-sector capacity development across dozens of countries.

Meanwhile, UNCTAD warned that although the world is expected to invest around $800 billion in AI infrastructure this year, most of that investment remains concentrated in a small number of countries. Developing economies, speakers argued, risk arriving ‘after the rules have already been written’ unless international cooperation accelerates.

The session concluded with broad agreement that the next phase of WSIS should focus less on developing new declarations and more on delivering measurable progress in connectivity, AI skills, trusted digital infrastructure, and inclusive governance.

Twenty years after the original WSIS process began, participants agreed that the challenge is no longer defining a vision for an inclusive information society, but ensuring that vision becomes reality.

Track all key moments from the WSIS Forum 2026 on our dedicated WSIS page.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot

OECD report warns AI skills gaps could widen labour inequalities

The OECD has warned that stronger skills policies will be needed to prevent AI from widening labour-market inequalities.

In its policy paper Skills in the AI age, the organisation says AI can boost productivity, support economic growth and create new opportunities. Still, it may also deepen existing gaps if workers and firms are not prepared for the transition.

AI adoption by firms has accelerated rapidly in OECD countries, rising from around 7% to 20% of businesses between 2021 and 2025.

The OECD says the increase has been driven partly by the spread of generative AI tools such as ChatGPT and Copilot.

Adoption remains uneven. Larger firms and start-ups are more likely to use AI, while small and medium-sized enterprises face barriers including costs, infrastructure gaps and shortages of skilled workers.

The paper also cautions that exposure to AI does not automatically mean a job is likely to disappear.

High-skill occupations such as managers, professionals and engineers are among the most exposed to AI, but are less likely to be automated because they rely heavily on non-routine cognitive and social skills.

Low- and middle-skill roles involving routine manual or cognitive tasks face higher automation risks.

The OECD says workers will need a mix of foundational skills, ICT skills and complementary skills such as critical thinking, creativity and collaboration.

Advanced AI skills, including machine learning and data science, remain scarce, with workers possessing such skills accounting for around 1% of the workforce.

The organisation calls for stronger education and training systems, wider lifelong learning, AI literacy for all workers, employer-led training and better coordination between governments, industry and education providers.

Why does it matter?

The OECD report frames AI skills as a core labour-market issue, not only a technology-sector concern. If training systems do not adapt, AI adoption could widen gaps between large firms and SMEs, between high- and low-skilled workers, and between regions with different levels of digital capacity. The report also makes an important distinction for policy: jobs highly exposed to AI are not necessarily the jobs most likely to disappear, meaning governments need more targeted approaches to reskilling, worker support and AI literacy.

Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!

NVIDIA says nations should build AI around local priorities

NVIDIA says countries are increasingly building AI around domestic infrastructure, local data, skilled workforces and national business ecosystems.

According to the company, this approach allows governments and industries to develop AI systems that reflect local languages, cultures, regulations and public priorities.

NVIDIA said national AI capabilities now go beyond computing infrastructure. Countries are also developing foundation models trained or fine-tuned on local datasets, helping systems better reflect regional dialects, cultural context and specific domains.

The company identifies five elements of a national AI strategy: trusted AI aligned with national goals, an AI-ready workforce, locally trained models and data, a strong domestic AI ecosystem and AI factories for training and inference.

NVIDIA describes AI factories as locally owned, operated and governed AI clouds that provide computing capacity through public-private partnerships.

The blog highlights examples, including AI agents supporting public-service workflows in France, multilingual AI models in India and AI tools for legal services in Brazil.

NVIDIA argues that domestic infrastructure, local data and homegrown talent can help countries apply AI to economic growth, public services, climate resilience, cybersecurity and social development.

Why does it matter?

NVIDIA’s framing reflects a broader shift in how governments and companies talk about AI: not only as a commercial technology, but as strategic infrastructure. Local compute, datasets, models and skills can help countries adapt AI to their own languages, laws and public needs. At the same time, the source is a vendor blog, so its emphasis on AI factories and accelerated computing should be read as part of NVIDIA’s commercial and policy positioning in the sovereign AI debate.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot

Viber brings ChatGPT into its messaging app

Rakuten Viber has launched ChatGPT-powered tools inside its messaging app through a new partnership with OpenAI.

The integration allows users to ask questions in a dedicated ChatGPT chat or tab, mention @ChatGPT in supported private and group chats, summarise conversations and shared links, polish draft messages, translate messages and remix images.

Viber said most tools are available after users update the app, without requiring ChatGPT registration.

Image Remix requires users to log in to ChatGPT within Viber or create a free account. OpenAI says availability may vary by region, app version, account and chat type.

The privacy model depends on the feature used. Viber says its core messaging features remain protected by end-to-end encryption, while ChatGPT-powered tools are activated only when users choose to use them.

When a ChatGPT-powered feature is used, Viber sends OpenAI the information needed to process that request. Depending on the feature, that may include selected messages, drafts, images, prompts, link content, messages that mention @ChatGPT, timestamps, approximate location and a Viber-generated hashed user ID.

OpenAI says data sent from ChatGPT-powered features in Viber personal and group chats is not used to train its models, except for conversations in the ChatGPT tab.

If a user connects a ChatGPT account, activity may be associated with that account and handled under OpenAI’s standard retention and data settings.

Why does it matter?

The launch brings generative AI into everyday messaging, moving ChatGPT from a separate assistant into conversations, links, drafts, translations and images. That makes AI tools more accessible, but also creates a more complex privacy model. Users need to understand when messages remain inside an end-to-end encrypted chat and when selected content is sent to OpenAI for processing. For messaging platforms, the key governance challenge is adding useful AI features while preserving user control, clear consent and transparent data handling.

Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!

Inaugural UN Global Dialogue on AI Governance ends with call to turn principles into action before 2027

The inaugural United Nations Global Dialogue on AI Governance concluded in Geneva with a clear message from governments, industry, civil society, and international organisations: the success of global AI governance will depend not on the principles adopted, but on the concrete actions taken before participants reconvene in New York in 2027. Speakers repeatedly argued that narrowing the widening AI divide, strengthening international cooperation, and embedding human rights into AI governance will require practical implementation rather than new declarations alone.

From principles to practice

Iceland’s President Halla Tómasdóttir opened the closing plenary by arguing that AI’s future will be shaped not by technological capability but by human choices about power, accountability, and inclusion. Drawing on Iceland’s experience of harnessing natural resources for the public good, she said AI should likewise serve society rather than narrow interests.

‘Access without agency is not inclusion,’ she said, warning that communities furthest from today’s centres of technological power must become co-authors of the AI future rather than passive recipients of it. She also cautioned that ‘principles without practice can inspire false comfort’, arguing that AI governance should ultimately be judged by whether it increases people’s dignity, opportunity, and hope.

Microsoft Vice Chair and President Brad Smith echoed those concerns, presenting new data showing that AI adoption has reached 27% of the working-age population in the Global North, compared with just 15% in the Global South. Without intervention, he warned, that gap is likely to widen further over the coming year. Smith identified four priorities for more equitable AI adoption: expanding access to electricity, completing global internet connectivity, supporting multilingual AI models, and making digital skills widely available. He also acknowledged that AI capabilities have advanced much faster than governance frameworks over the past year, calling interoperability between national governance approaches essential to avoid a fragmented regulatory landscape.

Building an inclusive global governance system

Several speakers underlined that no country or institution can govern AI alone and that the UN remains uniquely positioned to convene an inclusive international process.

Guy Ryder, UN Under-Secretary-General for Policy, described the organisation’s greatest strength as its ability to bring together all 193 UN member states alongside businesses, researchers, and civil society. He acknowledged the need for stronger coordination across UN agencies while arguing that the Global Dialogue should become a recurring platform connecting AI governance efforts across international organisations and forums.

Civil society representative Raman Jit Singh Chima of the Association for Progressive Communications urged policymakers to build on existing digital governance mechanisms such as the World Summit on the Information Society (WSIS) and the Internet Governance Forum (IGF), rather than replacing them with entirely new structures. He also warned that AI governance must be firmly grounded in human rights and informed by the experiences of women, girls, and marginalised communities, who are often disproportionately affected by AI systems while remaining underrepresented in governance discussions.

Namibia’s Minister of Information and Communications Technology Emma Theofelus shifted the discussion towards implementation, calling on the international community to help countries translate global AI principles into national legislation, invest in digital infrastructure, and strengthen scientific and technical capacity in developing economies. She argued that meaningful participation requires recognising countries’ different starting points rather than assuming all governments have equal resources and capabilities.

Dialogue identifies common priorities

The closing session also reflected on the discussions held across the Dialogue’s four thematic tracks, which collectively identified recurring governance gaps around infrastructure, funding, skills, trust, and participation.

Gaia Marcus, Director of the Ada Lovelace Institute, argued that public participation should become a source of evidence rather than a symbolic consultation exercise. Those most affected by AI systems, including workers facing automation and vulnerable communities, should have clear channels to influence policy decisions, she said, adding that trust depends on accountability rather than public relations.

Jaan Tallinn, co-founder of the Future of Life Institute, offered one of the session’s starkest warnings, arguing that AI capabilities are now advancing on a quarterly basis while governance processes typically evolve over years. He cautioned that frontier AI companies are pursuing increasingly capable systems despite acknowledging significant safety challenges, making international cooperation more urgent than ever.

Rapporteurs from the four thematic clusters highlighted broad convergence across the dialogue. Participants repeatedly called for stronger AI capacity building, practical interoperability between governance frameworks, greater support for local adaptation and multilingual AI, stronger human rights safeguards, and wider participation from developing countries and civil society. Rather than competing to build the most powerful AI systems, several speakers argued that countries should focus on developing the institutions, skills, and governance mechanisms needed to deploy AI responsibly.

Geneva lays the foundation for New York

Closing the event, co-chair Rein Tammsaar said the inaugural dialogue had brought together more than 4,200 registered participants from nearly 170 member states, alongside representatives of industry, academia, civil society, and international organisations. He argued that the discussions demonstrated the world’s challenge is no longer a lack of AI principles, but the absence of practical mechanisms to implement them.

Tammsaar said participants had moved beyond abstract debates towards discussions on national AI strategies, legal safeguards, teacher training, and public-sector capacity building. He also reiterated that the AI divide is about far more than access to technology, it is also about the ability to shape, govern, and benefit from AI.

Co-chair Egriselda López concluded that the Global Dialogue is intended not to replace existing AI governance initiatives but to connect them, strengthen cooperation, and help countries learn from one another. Recalling a remark from a representative of a small country who said, ‘We are here not to be a footnote,’ López said the statement captured the spirit of the inaugural dialogue.

Both co-chairs agreed that Geneva should be viewed not as the end of a process but as its beginning. They urged participants to return home with concrete commitments and practical actions so that, when the second UN Global Dialogue on AI Governance convenes in New York in May 2027, progress can be measured not by new principles but by tangible improvements in inclusion, capacity, and international cooperation.

Track all key moments from the Global Dialogue on AI Governance inaugural meeting on our dedicated page.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot

AI governance must serve all countries, ministers tell UN Global Dialogue

Ministers and senior officials from around the world used the third high-level governmental plenary of the UN Global Dialogue on AI Governance to outline national priorities for AI, while calling for stronger international cooperation to ensure AI benefits are shared more equitably. Although countries differed on regulatory approaches, participants broadly agreed that AI governance must be inclusive, human-centric and grounded in multilateral cooperation if it is to narrow rather than deepen global inequalities.

Throughout the session, speakers highlighted AI’s transformative potential for healthcare, education, agriculture and public services, while repeatedly warning that unequal access to computing power, infrastructure, talent and financing risks leaving many developing countries behind.

Europe pushes safety-by-design and evidence-based governance

Germany and the European Union placed safety, trust and evidence-based policymaking at the centre of their interventions.

Germany’s Federal Minister for Digital Transformation and Government Modernisation, Karsten Wildberger, described AI as ‘an entirely new paradigm’ developing at unprecedented speed and argued that governments must actively shape its future rather than react to it.

‘We must shape AI because otherwise AI will shape us,’ he said, urging countries to embed safety, security and respect for human values into AI systems from the outset instead of attempting to add safeguards later.

Wildberger also announced Germany’s new National AI Safety and Security Institute, which will evaluate advanced AI systems and contribute to international cooperation alongside industry, academia and civil society.

Representing the European Union, Director-General of DG CONNECT Roberto Viola similarly highlighted AI’s enormous promise, pointing to advances in biotechnology, drug discovery and robotics that could accelerate scientific progress and economic growth. At the same time, he warned that AI could also be used to manipulate children, attack critical infrastructure or amplify other societal risks if left without appropriate safeguards.

Viola stressed that governance should remain grounded in scientific evidence rather than political assumptions, praising the work of the Independent International Scientific Panel on AI as an important source of objective expertise for policymakers.

Developing countries call for ‘capacity before compliance’

If Europe focused on governance principles, developing countries focused on the practical barriers that prevent them from participating fully in the AI economy.

A recurring message throughout the plenary was that AI divides extend far beyond access to technology, encompassing shortages of computing power, electricity, broadband connectivity, quality datasets, skilled professionals and financial resources.

Indonesia’s Minister of Communications and Digital Affairs, Meutya Viada Hafid, argued that AI governance should support development rather than simply regulate risks. She introduced the principle of ‘capacity before compliance,’ warning that expecting countries with limited digital infrastructure to meet the same governance obligations as advanced AI economies would neither be realistic nor equitable.

Pakistan’s Minister of Information Technology and Telecommunication, Shaza Fatima Khawaja, similarly warned that the global ‘capability divide is real and it is widening.’ She urged countries to move beyond discussions of principles towards concrete investments in shared computing infrastructure, open-source models, regulatory sandboxes and a proposed global AI fund to help developing nations build sovereign AI capabilities.

Uganda highlighted that Africa currently possesses less than 1% of global AI computing capacity despite ambitious plans to use AI to support economic transformation, while Malawi described facing what it called an ‘impossible choice’ between accepting unacceptable risks or being left behind altogether.

Other speakers from Chad, Mozambique, Somalia and Mali echoed these concerns, arguing that AI governance should recognise different national circumstances while ensuring countries become active contributors to AI development rather than remaining dependent consumers of technologies designed elsewhere.

National strategies offer practical governance lessons

Alongside calls for greater international support, several governments presented national initiatives that they hope could contribute to future global governance models.

Thailand proposed serving as an international AI governance sandbox where global principles could be tested through practical implementation rather than remaining solely the subject of international discussions. Minister Chaichanok Chidchob warned that fragmented governance risks undermining trust itself and invited UN partners to develop scalable governance models through real-world experimentation.

Singapore shared lessons from its own AI governance experience, identifying reliable digital infrastructure, trusted access to high-quality data and strong public confidence as the three foundations of successful AI adoption. The country also highlighted its work on AI safety research and international technical standards.

Rwanda pointed to its national AI policy, newly established AI agency and broader Africa Declaration on Artificial Intelligence adopted by dozens of African countries as examples of regional cooperation designed to harmonise governance approaches.

Other governments showcased complementary initiatives. Chile proposed creating a multilateral network of AI sandboxes operating under common rules, while the Maldives argued that AI can only deliver meaningful public value when built on secure digital public infrastructure, trusted data systems and clear accountability mechanisms. Zimbabwe highlighted its recently launched National Artificial Intelligence Strategy and called for an international AI capacity-building fund alongside mutual recognition of AI ethics standards.

Interoperability emerges as a common governance goal

Although countries presented diverse national strategies, many converged around the idea that AI governance frameworks should be interoperable rather than identical.

Ireland argued that AI governance requires a shared international language built around transparency, accountability and human oversight, while the Netherlands described interoperability, not a single global rulebook, as the organising principle for the next phase of AI governance. Different jurisdictions, Dutch representatives argued, should be able to develop compatible systems based on common standards without sacrificing national flexibility.

Thailand echoed this concern, warning that fragmented governance could ultimately fragment trust itself. Indonesia similarly argued that trustworthy AI depends on interoperability rather than uniformity, while Singapore stressed the importance of internationally recognised technical standards that enable cooperation across borders.

This emphasis on compatibility reflected broader concerns that increasingly divergent national AI regulations could create unnecessary barriers to innovation, investment and international collaboration.

Cooperation remains the defining challenge

The closing interventions highlighted both the broad consensus and the remaining differences over how global AI governance should evolve.

India called on countries to choose ‘consensus over conflict’ before technological progress outpaces diplomacy, arguing that AI governance should provide every nation with a meaningful voice regardless of its level of technological development. Senegal promoted the Global Network for Cooperation on AI Capacity Building and welcomed proposals for a global AI fund to strengthen infrastructure and expertise in developing countries. Bahrain announced exploratory work on a potential global AI treaty initiative, while the United Kingdom highlighted partnerships helping countries across Africa develop local-language AI tools and strengthen domestic AI ecosystems.

The United States, meanwhile, emphasised voluntary cooperation with industry and a pro-innovation regulatory environment rather than binding international rules, illustrating one of the clearest policy differences to emerge during the session.

Despite these differing approaches, participants broadly agreed that AI’s future cannot be shaped by any country acting alone. As ministers repeatedly argued, the success of AI governance will ultimately be measured not by the sophistication of frontier models, but by whether countries of every size and level of development can safely use AI to improve the lives of their citizens.

Track all key moments from the Global Dialogue on AI Governance inaugural meeting on our dedicated page.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot