IGF Daily Summary
for Wednesday, 25 June 2025
Dear readers,
Welcome to our daily report from Day 2 at IGF 2025.
Day 2 of the IGF featured discussions across multiple sessions addressing AI governance, digital divides, cybersecurity, and the future of the IGF. Participants examined how emerging technologies are reshaping digital governance while exploring practical solutions for inclusion and protection online.
One specific issue raised in several AI discussions was the need for equitable access to AI computing power, particularly for the Global South. Key takeaways included the challenges of ‘compute deserts’ and the potential of regional consortia to pool investment and technical capacity. Speakers stressed the importance of not only hardware access but also digital skills development, AI literacy, and inclusive governance. Concrete projects, like leveraging AI and other technologies to transform Africa’s coffee value chain, showed how localised innovation can meet global goals. The tone was hopeful yet clear: collaboration, transparency, and political will are indispensable.
The WSIS+20 review process featured prominently, with participants calling for a permanent IGF mandate and sustainable funding, and noting the importance of an inclusive and transparent process, among other issues.
What stood out to you in yesterday’s discussions?
Diplo reporting team
Key questions from Day 2
How can we ensure AI development serves humanity rather than concentrating power?
Participants identified ‘a double concentration of power’ where only a ‘handful of private actors shape what people see, influence democratic debate, and dominate key markets without meaningful oversight’. The Freedom Online Coalition’s Joint Statement emphasised ‘putting humans at the centre of AI development’ with ‘clear obligations for both states and the private sector’. Technical solutions included ‘developing agent-to-agent interaction protocols and model context protocols to create interoperability among various agentic systems’ while policy approaches focused on ‘risk-based regulation where high-risk applications require high levels of safety and due diligence from providers’.
What does meaningful connectivity mean in practice?
Participants defined meaningful connectivity as ‘extending beyond broadband connection to include quality, affordability, digital skills, and capabilities such as financial transactions online, security awareness, and access to government services’. Device affordability emerged as the primary barrier, with ‘for the poorest 20% of people in sub-Saharan Africa, an entry-level internet-enabled device costs 99% of average monthly income’.
How do we protect children in an AI-enabled digital environment?
Participants emphasised that ‘the digital space, the internet was not designed for children’ and that ‘children are systematically exposed to risks and harms because services are designed with three main aims: increased time spent, increased reach, and increased engagement’. Solutions include ‘safety by design approaches’ and cross-platform collaboration. The Tech Coalition’s Lantern program was highlighted as the first cross-platform signal-sharing program that helps companies securely share signals about accounts and activity that violate child safety policies.
How do we bridge the compute divide in AI development?
Participants noted that ‘AI computing power is concentrated in roughly 30 nations, primarily the US and China’. To illustrate, Brazil has ‘1% of all data centres in the world’ and ‘according to EAA numbers, Brazil has 0.2% of computational power globally’. Solutions explored included ‘multistakeholder collaboration models similar to the Global Alliance for Vaccines and Immunization’ and ‘investment in local talent and support for community-driven research’.
Summary of discussions
AI governance and technological concentration
AI dominated governance discussions, with participants examining both opportunities and challenges. The High Level Session on AI and the Future of Work demonstrated AI’s transformative potential through concrete examples: agricultural extension services costs dropped from $35 per farmer to $0.3 per farmer through AI implementation, while Norway’s tax administration achieved 85% detection rates for missing tax returns using AI, compared to 12% before.
However, participants identified fundamental architectural differences between internet and AI systems. While the internet was built on ‘open, decentralised, transparent, interoperable architecture,’ AI represents ‘a highly centralised architecture’ that is ‘frequently proprietary, non-interoperable, and very opaque’.
The concentration of AI development emerged as a critical concern. Participants noted that AI computing power is concentrated in roughly 30 nations, primarily the US and China. In China alone, there are 434 large language models as of the session date.
Agentic AI systems present additional challenges. A Stanford study revealed that ‘almost half of the tasks by the AI agents’ described in the study ‘were seen by workers as non-desired for automation’. Bias issues were illustrated through a case from Argentina where an AI agent for career guidance ‘tended to recommend degrees that usually have lower tuition fees’ to low-income students, despite most Argentine universities not charging tuition fees.
Magical realism of AI
Like many technologies, AI is magical as it amplifies our capabilities and brings us beyond our limitations. Yet, the impact of AI on our lives is very realistic. The main challenge ahead of us is reconciling this ‘magical realism’ of AI and its governance.
Jovan Kurbalija, Main session on AI governance
Digital divides and meaningful connectivity
Digital inclusion discussions revealed persistent challenges despite decades of connectivity efforts. Like in previous days, participants noted that 2.6 billion people remain unconnected to the internet, but significantly, ‘only 4% of unconnected people live in areas without mobile broadband coverage’ while ‘90% of unconnected people live in covered areas but face other barriers’.
The concept of meaningful connectivity emerged as crucial, extending beyond broadband connection to encompass quality, affordability, digital skills, and broader capabilities, including people’s ability to transact financially online, awareness of security threats, data protection capabilities, and access to government services.
Device affordability was identified as the primary barrier. Participants noted that ‘for the poorest 20% of people in sub-Saharan Africa, an entry-level internet-enabled device costs 99% of average monthly income’.
Gender disparities persist, with ‘women in low and middle-income countries 14% less likely than men to use the internet’ and ‘this gender gap has stalled without significant progress’.
Cybersecurity and critical infrastructure protection
Cybersecurity discussions revealed an increasingly dangerous threat landscape. Participants noted that ‘nearly 40% of all documented cyber operations by states in 2024 focused on critical infrastructure’ while ‘ransomware attacks surged by 275%.’ The scale was illustrated by the observation that if cybercrime were a country measured by GDP, it would be the world’s third largest economy.
The interconnected nature of infrastructure creates cascading vulnerabilities where ‘a compromise in one sector, say electricity, can ripple into others like healthcare, telecommunications or transportation.’
Information sharing is both crucial and challenging. While its value for digital resilience is widely recognised, participants noted that ‘in practice, it proves to be something that is actually very difficult’ because intelligence agencies emphasise secrecy and have limited incentive to share; companies face reputational damage and liability risks when sharing breach information; and threat intelligence companies tend to protect their business models.
The most effective information sharers were identified as CISRTs and the CERT community which ‘have a long-standing tradition in the international community of exchanging information, and they see it as a core value’.
Child protection online is a critical priority. Participants noted that ‘roughly 300 million victims of child sexual abuse and exploitation every year globally; that’s about 14% of the world’s children each year’. The challenge is compounded because ‘bad actors typically exploit multiple services across the tech ecosystem in their attempts to groom children, distribute CSAM, or engage in other harmful activities like financial extortion’. Mentioned solutions included cross-platform collaboration through initiatives like the Tech Coalition’s Lantern programme which was highlighted as the ‘first cross-platform signal sharing programme’ that helps companies securely share signals about accounts and activity that violate child safety policies.
Content governance and AI-generated misinformation
The challenge of combating misinformation intensified with AI-enabled content creation. Participants documented dramatic growth of AI content forums monitored grew from 49 sites in May 2023 to 1,271 sites currently with ‘one individual behind more than 273 AI-generated websites imitating local news sites in the US and Germany.
AI reliability emerged as a critical concern, with research showing that AI chatbots repeat false claims authoritatively approximately 26% of the time when tested, creating what participants described as ‘a vicious circle of disinformation.’
Sexual deepfakes affecting teenagers became a particular focus. In Korea, policy reports about deepfake sex crimes increased from 156 cases in 2021 to 1,202 cases in 2024. The availability of creation tools was highlighted as problematic, with ‘nearly 35,000 AI models available for public download on one platform service for generative AI, many of which are even marketed or with the intention to generate NCIIs, non-consensual intimate imagery’..’ Educational responses showed promise. The Korean Ministry of Education, for instance, published five guidebooks tailored to different age groups covering three key situations: being a victim, witnessing someone else as a victim, and causing harm.
WSIS+20 and the IGF’s institutional evolution
The WSIS+20 review process dominated institutional discussions. Multiple participants called for a permanent mandate for the IGF with sustainable funding. Co-facilitators noted receiving ‘calls for either renewal of the mandate of the IGF, or a strengthening of that mandate, or even in some cases, calls to make it a permanent institution that’s able to access financing from regular resources of the United Nations’.
Participants emphasised the IGF’s unique role as ‘the only place across all the fora that we have about digital transition where all stakeholders are on equal footing’. This is especially important as some strategic foresight exercises reveal the concerning trend of ‘multistakeholder processes either being hollowed out or kind of completely undermined by corporate actors and state actors’.
Some points were made about the ability of the IGF to have its outcomes and outputs conveyed to decision-makers. Participants noted that ‘limited by its design, a lot of the great IGF discussions and outcomes do not necessarily land in decision-making fora at the UN regional or national levels’.
Despite challenges, participants proposed that a redesigned IGF, ‘a redesigned and a braver IGF, redesigned in terms of making it much more participative and innovative, in terms of the methodologies we use for our sessions, and a braver IGF, more willing to actually ask difficult questions around which there’s not going to be consensus.’
SDGs in focus
Digital technologies’ role in advancing SDGs featured in discussions, though progress remains uneven. Participants noted that with only 17% of SDGs on target, there is recognition of gaps still experienced five years into 2030, which is the target date for sustainable development goals.
Research showed that ‘digital technologies directly benefit 70% of SDG targets’, with strong correlations found between the ICT development index and SDG index, particularly on economic development. Specific examples included SDG 1 (no poverty), showing broadband expansion correlating with lower poverty, SDG 3 (good health and well-being), with universal health coverage correlating with ICT development, and SDG 4 (quality education), showing literacy and enrollment rates with positive trends related to higher ICT development index scores.
Digital public infrastructure was cited as having ‘immense potential’, with Harvard University estimates suggesting it can unlock value equivalent to 3 to 13 per cent of GDP, with an average improvement of 6 per cent for emerging economies.
However, funding gaps persist. Participants noted that ‘only 3% of official development assistance is going toward digitalisation, and digital public goods are being underfunded and undervalued’.
The African context highlighted particular challenges, with participants noting that ‘500 million of our population without any legal form of identity’ and ‘700 million of our population don’t have access to electricity’.
Thought-provoking ideas and questions
‘What economic incentive do people have to be creative, to do great things, to work hard?’ This question was raised in the context of concerns about AI companies using human-created content without compensation.
‘If we’re so clever with technology, why can’t we make something that, when once we’ve put an image online, it becomes indelible, it becomes unchangeable?’ This question was raised in the context of combating sexual deepfakes.
The IGF we want
– The IGF mandate should be made permanent with stable and predictable funding
– Consideration could be given to rebranding the IGF as a Digital Governance Forum, to more accurately reflect the fact that it has expanded its scope beyond internet governance to include AI and other emerging technologies
– Better integration between the IGF and NRIs, so that contributions from local communities are better brought to the global IGF
– A redesigned IGF that is ‘much more participative and innovative, in terms of the methodologies we use for our sessions, and a braver IGF, more willing to actually ask difficult questions around which there’s not going to be consensus’
– Dynamic Coalitions should be integrated into the main program rather than treated as side activities
Diplo/GIP at IGF2025
Diplo is partnering with the IGF Secretariat and the Government of Norway (as host country) to deliver AI-enabled, just-in-time reporting from the IGF 2025 meeting. Building on a decade of just-in-time IGF reporting, we will continue to provide timely and comprehensive coverage from the forum. Our reporting initiative will include session reports, an ‘Ask IGF 2025’ AI assistant, daily highlights, and more, available on our dedicated IGF 2025 web page on the Digital Watch Observatory.
Diplo’s Executive Director, Jovan Kurbalija, launched the eighth edition of his seminal textbook ‘Introduction to Internet Governance’, marking a return to writing after a nine-year pause. The session unpacked not just the content of the new edition but also the reasoning behind retaining its original title in an era buzzing with buzzwords like ‘AI governance’ and ‘digital governance.’
Kurbalija was also featured as a speaker during Wednesday’s main session on the governance of AI, noting that the policy community should make AI governance common sense, bottom-up, and explainable to anyone who is using AI.
What to expect?
On Thursday, 26 June, the Diplo-led CADE consortium will organise an in-person lightning session to explore multilingual challenges in global processes. The session will recreate a ‘Tower of Babel’ experience, inviting participants to speak in their native languages and attempt spontaneous communication. The activity will be followed by a group discussion on real-world solutions to language barriers, focusing on adaptation, inclusion, and communication accommodations.
At IGF Lillestrøm? Find us at the Diplo and GIP booth (#45) and CADE booth (#57) – also online in the virtual village.
Do you like what you’re reading? Bookmark us at https://dig.watch/event/internet-governance-forum-2025 and tweet us @DigWatchWorld
Have you heard something new during the discussions, but we’ve missed it? Send us your suggestions at digitalwatch@diplomacy.edu.