Developments in the internet governance environment July to September 2026

Wolfgang Kleinwächter Oct 21, 2026 35 min read

In the third quarter of 2026, the debate on Internet governance was dominated by three topics:

IGF, GDC and the new WSIS architecture
Artificial Intelligence
Cybersecurity

I. IGF and the new WSIS architecture

The most significant events in the third quarter were the preparations for the 21st IGF in Nairobi, in particular the MAG consultations on 6 September 2026 in Geneva, the WSIS Forum from 6 to 10 July 2026 in Geneva, and the start of the review of the Global Digital Compact (GDC) with ‘Digital Development Day’ on 21 September 2026 in New York.

IGF restructuring: new policy networks and preparations for Nairobi

With the new Chair of the IGF’s Multistakeholder Advisory Group (MAG), Jennifer Chung, a new dynamic has been injected into the IGF. The MAG previously met twice a year for open consultations in Geneva. It has now switched to a fortnightly schedule and meets online[1]. These virtual meetings are also open to the broader public. In addition, an ‘Office Hour’ with the MAG Chair has been established.[2] Stakeholders can now contact MAG members directly every two weeks to raise their concerns. Among the most important decisions taken between July and September 2026 were the restructuring of the ‘Intersessional Work’ and the overhaul of the programme for the IGF in December 2026.

The restructuring of ‘Intersessional Work’ focused primarily on refining the work of the Policy Networks, discussing the European Commission’s proposal on IGF Labs, realigning the MAG Working Group Strategy, the work of the Dynamic Coalitions, and closer coordination with the national and regional IGFs (NRIs).

The main work between the annual IGFs will in future take place within the four Policy Networks (1. Open Digital Infrastructure; 2. Digital Inclusion; 3. cyber security and resilience; 4. AI). They will already be submitting reports and working papers containing recommendations for the IGF in Nairobi. In accordance with paragraph 100 of the WSIS+20 Output Document, these are to be forwarded to other UN organisations. These organisations, in turn, are expected to play an active role in the discussions at the IGF in Nairobi. This is intended to contribute overall to enhancing the political relevance of the IGF. In the medium term, the Policy Networks could develop into a kind of independent IGF sub-organisation.
The MAG Working Group on Strategy has been reorganised. The new co-chairs are MAG members Olaf Kolkman (ISOC), Peace Oliver Amuge (APC) and Rajendra Pratap Gupta (India). The Working Group’s mandate „to support the MAG in developing actions, implementable in the short and medium term, to strengthen the IGF and position it strategically in the evolving digital cooperation landscape“[3] has been confirmed. The Strategy Working Group will meet every three weeks in future.[4]
The proposal to establish IGF-Labs came from the European Commission. IGF Labs are intended to function as early-warning systems. The aim is to explore governance proposals relating to emerging technologies for an interoperable, free and open Internet before they become entrenched in technical architectures or regulatory approaches. Conceptually, much remains unclear, particularly how the IGF Labs will fit into the existing IGF architecture comprising PNs, DCs and NRIs. By early September 2026, the IGF Secretariat had received 124 proposals for IGF Labs. The proposals will be evaluated by mid-October 2026. The MAG intends to select no more than three IGF Labs from the proposals as pilot projects.

Preparations for the 21st IGF in 2026 made good progress in the third quarter of 2026. The key elements of the programme, comprising the main sessions and workshops, are now in place. Planning for the special sessions – Governmental Dialogue, Parliamentary Track, Judiciary Engagement, Youth Track and Business Engagement – is nearing completion. The venue for the IGF will not be the UN Office in Nairobi, where the 11th IGF took place in 2011, but the newly opened Bomas International Convention Centre.[5] The host expects more than 10,000 participants.

There were 456 submissions in response to the call workshops published in July 2026. At the MAG consultations on 7 September 2026, agreement was reached on 80 workshops, grouped around five themes (1. Economy, Innovation & Sustainability; 2. Governance, Cooperation & Institutions; 3. Infrastructure, Access & Technical Foundations; 4. Rights, Inclusion & Human Developmentn und 5. Trust, Safety & Information Integrity).[6]
As regards the plenary sessions (Main Sessions), four topics were agreed upon at the virtual MAG meeting on 23 September 2026: 1. AI governance; 2. social media, online harms and digital safety; 3. digital infrastructure and commons; und 4. digital inclusion and cooperation. The outlines of the Main Sessions will be published on 18 October 2026.[7]
The expected outcome of the IGF in Nairobi should be as practical as possible and made manageable for other policy processes. „Short thematic briefs, concise session-level takeaways and formats that can be submitted directly to relevant processes were generally preferred over a single long report“.[8] This also involves closer integration with ongoing intergovernmental negotiations in the areas of cybersecurity, human rights, digital trade and AI. For example, the PN AI is to address the outcomes of the Global Dialogue on AI Governance. The proposal made in the EURODIG Messages of June 2026 (Brussels) that the exchange between the UN AI Panel and the IGF MAG Liaisons has not yet been discussed.
The ‘Governmental Dialogue’, newely incorporated into the IGF under paragraph 101 of the WSIS+20 Output Document, is currently being prepared by a working group (comprising Kenya, Egypt, Australia, the United Kingdom, Pakistan und Switzerland). The role of governments within the IGF was one of the most contentious issues during the WSIS+20 Review in December 2025.

New WSIS architecture: focus on action lines and strengthening UNGIS

The WSIS Forum 2026 took place in Geneva from 6 to 10 July 2026.[9] It was the first meeting following the WSIS+20 review by the UN General Assembly. More than 1,000 speakers, including 78 ministers, took part in over 250 sessions. The WSIS Forum was part of ‘Digital Geneva Week’ and took place alongside the UN’s ‘Global Dialogue on AI Governance’ and the ITU’s ‘AI for Good Summit’. In total, the three events attracted more than 12,000 participants from 177 countries.

Unlike the IGF, the WSIS Forum focuses on implementing the objectives of the eleven WSIS action lines reaffirmed at WSIS+20. With this new mandate, work began on roadmaps leading up to 2035 (WSIS+30). The dominant call was to move from resolutions to action – that is, to measurable outcomes. In terms of content, the focus was on digital inclusion, bridging the digital divide, AI and cyber security.

On the subject of digital inclusion, it was emphasised that mere connectivity is not enough for the Global South. What is needed is affordable access, digital skills, stable and reliable legal frameworks, the protection of women and girls online, and genuine opportunities for young people to have their say.
On the subject of AI the discussion centred on global guidelines, reliable testing, risk standards, the new ‘AI divide’, the dangers of technology being monopolised by a handful of companies and states, and the need to avoid fragmentation in global discussions on AI governance.

On the sidelines of the WSIS Forum, a new ‘AI for Good Global Commission’ (AIGGC) was established on 7 July 2026.[10] It is chaired by Paul Kagane, President of Rwanda, and Marc Benioff, the multi-billionaire CEO of Salesforce. The 40 members include high-ranking politicians, amongst them the Presidents of Estonia and Iceland, Directors-General of UN organisations, as well as the CEOs of Nvidia, Mistral, Amazon and Anthropic, and the President of Microsoft, Brad Smith. Civil society is represented by just one member: the Director of APC. There is no German member on the commission. The AIGGC replaces the ‘Broadband Commission for Sustainable Development’, established over ten years ago by the ITU and UNESCO. The Broadband Commission had published annual reports on progress towards achieving global interconnectivity. It remains unclear how this new commission will liaise with the UN’s AI panel and the Global Dialogue on AI Governance.
The first meeting of the newly constituted UN Group for the Information Society (UNGIS) took place as part of the WSIS Forum[11]. UNGIS’s status had been strengthened by the WSIS+20 outcome document, partly to balance the new, prominent role of ODET (UN Office of Emerging Technologies) in New York. UNGIS is a network of intergovernmental organisations. It has 31 members, primarily UN organisations, and 21 observers. The UNGIS meeting adopted a work plan for 2026/2027. The aim is to improve coordination amongst UN organisations in implementing the WSIS action lines. Duplication is to be avoided and greater output achieved. The UNGIS chairmanship passed from UNCTAD to the ITU. Through the WSIS+20 document, UNGIS has been given an additional mandate to address the financing of digital development and to improve cooperation with non-state stakeholders.

In accordance with paragraph 67 of the WSIS+20 resolution, a ‘UNGIS Task Force on Strengthening Financial Mechanisms for Digital Development’ was established. It must submit a report containing concrete proposals by summer 2027. The Task Force comprises 29 members from various UN organisations, including the World Bank and the World Trade Organisation (WTO)[12]. The first meeting took place on 8 September 2026 in Geneva.[13]
Paragraph 121 called on UNGIS to involve non-state stakeholders more closely. To this end, a ‘UNGIS Multistakeholder Consultation Group’(MCG) was established with the mandate ‘to serve as an advisory and consultative body’, but without voting rights on decisions.[14] Over 100 nominations have been received for the MCG. The composition is due to be announced in mid-October 2026. A virtual kick-off meeting is planned for 28 October 2026.

Global Digital Compact (GDC): Launch of the review and first reports from the UN AI Panel

The review of the ‘Global Digital Compact’, adopted by the UN General Assembly in 2024, is on the agenda for 2027. The ‘Digital Development Day’ held in New York on 21 September 2026[15] marked the first step in preparing for this review. The focus was on an initial assessment of the GDC’s resolutions on AI and data governance.

The AI panel began its work in April 2026 with a first meeting in Madrid, presenting a ‘Preliminary Report’ in early July 2026 and an initial ‘Thematic Brief’ in September 2026. A second meeting is scheduled for November 2026 in Barbados. The ‘Global Dialogue on AI Governance’, also adopted by the GDC, took place in Geneva from 7 to 8 July 2026. The second dialogue is scheduled for 3 and 4 May 2027 in New York.
The ‘Working Group on Data Governance at all Levels’ (WGDG), meeting under the auspices of the UNCSTD, held its seventh session on 8 and 9 September 2026 in Geneva.[16] Discussions centred on elements of the final report, which must be submitted to the 81st UN General Assembly by the end of the year. As the group’s mandate is very broad, the interim findings to date are also rather vague and barely go beyond platitudes. According to the report, ‘data governance’ forms the basis for AI governance. ‘Data governance’ should therefore be based on nine principles (1. Accountability, 2. Transparency, 3. Security, 4. Privacy, 5. Data Quality, 6. Interoperability, 7. Equity, 8. Equitable Development, 9. Multistakeholder Participation). These principles are essentially derived from the Internet governance principles. Data governance is to be regarded as a two-layer mechanism (1. Universal Principles, applicable across all jurisdictions; 2. Sector-Specific Adaptation, proportional to risk, sensitivity and context). The group comprises over 50 members from all stakeholder groups. The final meeting is scheduled for 4 and 5 November 2026 in Geneva.
It is expected that the 81st UN General Assembly will adopt a resolution in December 2026 setting out the modalities and roadmap for the GDC review. As with the WSIS+20 review, it can be assumed that two co-facilitators will be appointed to lead a process of multi-stakeholder consultations and intergovernmental negotiations, which will ultimately result in a decision on the future of the GDC to be adopted by the 82nd UN General Assembly in December 2027.

II. AI Governance

The issue of AI governance is increasingly overlapping with the debate on internet governance and is becoming a key topic in global politics. This is driven primarily by rapid technological development. The degree of autonomy in new AI applications is constantly growing and is on the verge of crossing a critical threshold. Individual applications are beginning to slip beyond human control.

Growing risks and the US debate: hacks, calls for regulation and ‘superintelligence’

The autonomous hacks by AI agents from OpenAI, Anthropic, Google, Meta and others, which came to light in the third quarter, have led to new, incalculable risks, triggering alarmist warnings from AI insiders and fuelling calls for regulation and a new global AI policy. In the US in particular, the debate has escalated into a media frenzy. In Europe, these risk concerns are generally shared; however, Arthur Mersch, CEO of Mistral, Europe’s largest AI company, also sees this as a strategy by US companies to entrench the ‘ ’ status quo, thereby securing the US’s global lead.[17] On 15 September 2026, China’s Minister of State Security, Chen Yixin, also warned of AI risks that threaten national security.[18]

On 26 August 2026, Bill Gates, founder and former CEO of Microsoft, issued a statement in which he warned of the risks posed by AI. “Artificial intelligence (AI) will either be the greatest equaliser of all time or the worst source of injustice. AI-fuelled fraud, disinformation, deepfakes and surveillance are the dangers that many people will feel most acutely in their everyday lives. AI capabilities are already being misused for cyber attacks. The very same AI model that can detect software bugs and enable companies to fix them can also help criminals to exploit them. Think of the infrastructure that will be put at risk as a result: hospitals, financial institutions, water supply systems, electricity grids, and systems for administering state benefits. If these institutions are attacked, it is the patients, customers and benefit recipients who will bear the brunt of it.”[19] The world needs a plan, argues Gates: “This unprecedented technology demands an unprecedented global response.”
In early September 2026, Dario Amodei, CEO of Anthropic, published an essay entitled “We Must Pace the Frontier”.[20] Amodei sees risks primarily in “loss of control over AI systems, the misuse of AI for cyber attacks and bioterrorism, and serious economic disruption”. Altmann has put forward several proposals to manage these risks: “Embedded evaluators. Democratic coordination, global coordination. Operational excellence, monitoring, sandboxing, training, environment hygiene, alignment. Interpretability and testing”. He calls for smart regulation, but points out that regulators are generally slower than developers, particularly when it comes to international rules. Amodei’s proposal was met with widespread acclaim and received support from Anthropic’s competitors, such as Elon Musk and Sam Altmann of OpenAI.
As early as 28 July 2026, over 1,000 AI experts and employees of major US AI companies had addressed the US government with an appeal entitled ‘Pacing the Frontier’[1] : ‘We request that the US government support an international effort to develop the technical and governance tools needed to deliberately pace the frontier of automated AI development.’[22]
The debate was fuelled by the new book by Eliezer Yudkowsky and Nate Soares, *If Anyone Builds It, Everyone Dies: Why Superhuman AI Would Kill Us All*, in which the authors describe a scenario leading to the end of humanity[23].
However, the ‘smart regulation’ called for by businesses fell on deaf ears with the US administration. US President Trump argued that regulatory barriers would be counterproductive for the US in winning the competition with China, and rejected any form of government regulation or curbing of AI development.[24] Speaking before the UN General Assembly on 22 September 2026, Trump said: “Whoever wins AI – you have to remember this – and now I say, whoever wins SI, whoever wins super intelligence wins. That’s the group that wins.”[25]
At a summit with leading AI companies in Washington on 29 September 2026, Trump and five of the US’s largest AI companies agreed on a mechanism for AI self-regulation. The ‘White House Accord on Super Intelligence’[26] contains four points. “1. We will implement robust internal controls to monitor the performance and alignment of our models during training and deployment in areas such as cyber security, biosecurity and chemical threats, and to ensure that our models do not inadvertently access or hack into technical systems. 2. We will empower an internal team to ensure the proper functioning of all controls, monitoring and detection mechanisms. 3. We will work with an independent external auditor or assessor to carry out independent assessments. 4. We will appoint an independent committee of the Supervisory Board to oversee the teams implementing the controls, as well as the internal and external auditors and assessors, and to receive their reports.” The meeting was attended by Elon Musk, Mark Zuckerberg (Meta), Jeff Bezos (Amazon), Jensen Huang (Nvidia), Dario Amodei (Anthopic), Sundar Pichai (Google) and Alex Carp (Palantir). The ‘White House Accord’ was signed by Trump, Musk, Zuckerberg, Huang, Amodei, Pichai and OpenAI President Greg Brockman.
Also on 29 September 2026, President Trump issued an executive order entitled ‘Inaugurating the Era of Super Intelligence’ (SI). Under this executive order, the term ‘Artificial Intelligence’ (AI), which had been in use until then, is to be replaced by the term ‘Super Intelligence’ (SI).[27] Trump had already announced this renaming during his speech to the 81st UN General Assembly on 22 September 2026 in New York, where he described AI as ‘fake news’. [28] Michael Kraitsos, White House Director of Science and Technology, was tasked with proposing a legally valid definition of ‘Super Intelligence’ within the next 60 days.

It remains unclear what consequences this will have for established AI institutions both within and outside the UN. However, the new terminology met with resistance even within the US. The Democratic Governor of California, Gavin Newsom, rejected the new terminology. “What I heard yesterday in Washingtin D.C. is bullshit and that should scared the hell out of everybody.“[29] On 30 September 2026, Newsom signed a California State Executive Order stipulating that all organisations within California must use the term “Artificial Intelligence”, regardless of decisions made by the “federal government”[30] California is home to the largest AI companies in the US. It remains to be seen whether this ‘war of words’ masks a ‘war of concepts’. It also remains to be seen whether the AI companies in Silicon Valley will follow the Californian directives or the White House regulations.
One side effect of Trump’s name change was a rush to register .si domains - the ccTLD Top Level Domain of Slovenia. In September 2026 alone, there were over 50,000 new registrations, bringing the total number of .si domains from 200,000 to 250,000.[31]

Democratic Senator Richard Blumenthal described the ‘White House Accord’ as dangerous and pointless. Self-regulation does not work, as the debate on the negative consequences of social media has shown.[32] The agreement amounts to ‘no public accountability, completely discretionary and a false sense of security’[33]
On 4 October 2026, Trump took his next step. He announced the establishment of a “Super Intelligence Force” (SIF). The SIF is set to become the central AI coordination body within the US government and “to ensure that America continues to lead the world in super intelligence, which many say is bigger than the Industrial Revolution and the Internet”. The SIF is headed by the former intelligence coordinator, Jay Clayton. He is joined by three deputies, including the Deputy Secretary of War, Emil Michael. An external advisory board includes, amongst others, former Secretary of State Condoleezza Rice.[34]
On 5 October 2026, the CEO of Mistral, Arthur Mersch, gave an interview to the *Süddeutsche Zeitung* in which he refuted the “doomsday scenarios” put forward by American AI entrepreneurs in the US. “There is no doomsday scenario. These are software systems, and they can be switched off. There is a strong tendency to anthropomorphise AI agents, as if they had a will of their own. That is not true. It is software that is dynamic and can take us by surprise. But you can switch it off in a second.” The systems are becoming increasingly autonomous. That is why we need safeguards, i.e. ‘monitoring systems’ (sandboxes) and ‘training systems’ (alignment). Mersch described the call by OpenAI and Anthropic for a pause in AI development as ‘regulatory capture’, the aim of which is to cement their dominance.[35]
On 14 September 2026, Chen Yixin, China’s Minister of State Security, published an article in the journal *China Cyberspace* in which he warned that AI applications could undermine China’s national security: “Various hostile forces and malicious individuals could misuse generative AI technologies – such as deepfake audio and video recordings, AI-generated images and text, as well as intelligent online armies – to produce political rumours on a massive scale.” These tools could “spread harmful information, stir up hostile sentiments, and wage a war for the broader public and cognitive warfare.” Yixin cited AI-driven attacks on infrastructure, data theft and fundamental changes in the conduct of warfare as further threats. [36]

An overview of international diplomacy

The new nature of the debate on AI governance is also reflected at the international level, for example in US-China relations, at the UN, in numerous regional intergovernmental and non-governmental bodies, and in the domestic policies of many countries.

At the summit between the Presidents of China and the US, Xi and Trump, held in Washington on 24 September 2026, AI was a key topic. As a result, both sides agreed to establish a communication channel for serious AI incidents, modelled on the ‘hotline’ between the Soviet Union and the US designed to prevent nuclear incidents in the 1960s, as well as to launch a Sino-American AI dialogue.
On 23 September 2026, the UN Security Council held its first-ever special session on AI. At the 81st UN General Assembly, which began on 22 September 2026, numerous heads of state and government spoke on AI policy. In July 2026, the new ‘Global Dialogue on AI Governance’ in Geneva had discussed the first ‘Preliminary Report’ from the new UN Panel on AI. The discussion continued at the ‘Digital Cooperation Day’ in New York on 21 September 2026, where a new ‘Thematic Brief’ from the UN Panel on AI regarding the ‘OpenAI vs. Huggling Face’ incident was published.
AI was on the agenda at the summits of the Shanghai Cooperation Organisation (SCO) in Bishkek and the BRICS states in New Delhi, both of which took place in early September 2026.
Other key developments in AI governance during the third quarter of 2026 included the founding of the ‘World Artificial Intelligence Cooperation Organisation’ (WAICO) on 23 July 2026 in Shanghai, the entry into force of the EU’s AI Act, together with the Digital Omnibus, on 1 August 2026, the adoption of the ‘Caroline Principles for Emerging Technologies’ by the G20 digital ministers on 2 September 2026 in North Carolina, and the ‘Call for Control of Frontier AI Models’ initiated by the Prime Ministers of Finland and Norway on 23 September 2026, which was signed by 22 states, including Federal Chancellor Friedrich Merz.

US-China Summit: AI at the heart of US-China relations

AI was the focus of the US-China summit held in Washington on 23 September 2026. The issue is a top priority for both sides, but they pursue different objectives and employ different methods.

As announced in their action plan ‘Winning the Race’ of 10 July 2025[37], the US aims to remain the world’s leading AI power and to set the global ‘gold standard’ for AI applications. They reject global agreements on AI governance within the framework of the UN and are committed to ‘permissionless innovation’. As an international instrument to promote its objectives, the US launched the ‘Pax Silica’[38] on 11 December 2025, which is supported by 24 states, including Germany.
China is likewise striving for a leading role in AI development but, as announced in the ‘Global AI Governance Action Plan’ of 25 July 2025[39], places greater emphasis on cooperation both within and outside the UN, particularly with the Global South. China places greater emphasis on the risks and insists that AI development must not be allowed to run its own course but must remain under human control, whilst also advocating for as few restrictions on innovation as possible. As an international instrument to underpin its position, China established a new intergovernmental AI organisation (World Artificial Intelligence Cooperation Organisation/WAICO)[40] in Shanghai on 23 July 2026, to which 36 states have now acceded. It is intended to be guided by the principles of the UN.
Regardless of the contradictions and conflicting strategies, both presidents sought potential common ground at the summit in Washington in September 2026. They see this primarily in areas where common threats could arise. This is reflected in the agreement to establish a ‘hotline’ for AI incidents and the creation of a new ‘US-China Super Intelligence (SI) Dialogue’.[41] The dialogue is due to convene for the first time in November 2026. Details on how the new AI crisis communication channel will operate are not yet known. Incidentally, the White House statement mentions that both sides had agreed on the term ‘Super Intelligence’ for the dialogue.[42] The Chinese side, however, continues to refer to ‘Artificial Intelligence’.[43]

Special historic session: The UN Security Council debates AI – and internet governance

On 23 September 2026, the UN Security Council addressed the topic of AI for the first time[44]. Invited as external experts were Joshua Bengio, co-chair of the UN’s AI panel and professor at the University of Montreal, as well as the CEOs of OpenAI, Sam Altman; Anthropic, Damio Amodei; and Hugging Face, Clément Delangue. Hugging Face was the victim of an autonomous hacker attack carried out by an AI developed by OpenAI.

Joshua Bengio compared current AI development to a car driving at high speed into a bank of fog, with children and grandchildren in the back seat. “A mother in Shanghai wants her sons to grow up in a world where AI improves their lives, not makes them worse. A father in Chicago wants his daughters to flourish thanks to AI, not suffer from its consequences. Our hopes and concerns are the same in every country. Humanity has previously demonstrated the courage and wisdom to manage catastrophic risks. This is one of those moments; it is far more important than any personal, economic or national interest.”[45] He called on the international community to agree on universal rules (guardrails).
The CEOs of OpenAI and Anthropic also spoke of a turning point. “If AI is to be democratic, the most important decisions must not be made solely by laboratories in San Francisco. They must be shaped through democratic processes and by governments that are accountable to the people they serve,”[46] said Sam Altmann. “We need a mechanism for complementary national and international standards for frontier AI: standards to measure capabilities, assess risks, determine the adequacy of safeguards and ensure effective human oversight as systems become more autonomous. We need common standards so that countries can compare findings, verify compliance and develop a common language and a shared understanding of what is happening.”
In the ensuing debate, the overwhelming majority of UN Security Council members called for concrete measures to bring AI risks under control, a proposal rejected by the US. Michael Kraitsos, Director of New Technologies at the White House, said: “The US totally rejects any attempt to construct a globalist scheme of control for superintelligence”. He saw no reason to pause AI development. As the sole technological superpower, the US is in the process of establishing the ‘gold standard’ for safe AI and is prepared to share this with other like-minded nations via the US export programme. ‘A prosperous future will not be secured by a global regulator. It will be secured by sovereign nations that adopt superintelligence, responsible companies that build it, and free people who refuse to be ruled by fear.’[47] He also used the term ‘superintelligence, formerly known as AI’.
Russia’s representative, Alexander Repkin, stated that the UN Security Council had no jurisdiction over AI security. This should be discussed within the new Global Mechanism for Cybersecurity (G-MEC) or in the negotiations on autonomous weapon systems (GGE LAWS)[48].
None of the speakers commented on the relationship between AI governance and internet governance. The UNCSTD Working Group on Data Governance (WGDG) has identified ‘data governance’ as the basis for ‘AI governance’ and, in its proposed principles for data governance, has drawn on the principles recognised for internet governance.

It appears that a new community is emerging around AI governance, one that has few links to the Internet community that has been taking shape since the late 1990s. Among the 40 members of the UN panel on AI, with the exception of co-chair Maria Ressa – who was co-chair of the IGF Leadership Panel – there is not a single expert with roots in the Internet community. But even Maria Ressa, a journalist and winner of the 2022 Nobel Peace Prize, joined the IGF Leadership Panel as an outsider. The CEOs of the leading AI companies also come from a different community to the founding fathers of the Internet and represent a new generation.
The debate surrounding internet governance, data governance and AI governance is reminiscent of the debate over the term ‘information society’ in the early 2000s. Alternatives at the time were ‘network society’ or ‘knowledge society’. It was argued at the time that ‘data’ was the raw material that becomes ‘information’ and is the prerequisite for ‘knowledge’. In today’s academic debate, the question of an ‘AI Society’ is being discussed, fuelled by publications such as Alexander Klimburgs's ‘The Darkening Web: The War for Cyberspace’ from 2018[49], ‘The AI Republic: Building the Nexus Between Humans and Intelligent Automation’[50] by Terence Tse, Mark Esposito and Danny Goh from 2019, or ‘The Technological Republic: Hard Power, Soft Belief, and the Future of the West’[51] by Alexander C. Karp, CEO of Palantir, and Nicholas W. Zamiska from 2025.

Global forums: the UN General Assembly, the G20 and the European AI strategy

During the general debate of the 81st UN General Assembly, which began on 21 September 2026 in New York, 128 states spoke on the topic of AI. Four years ago, the topic was raised in only four interventions.

Right at the opening, UN Secretary-General Antonio Guterres had already identified AI as one of the most pressing issues of our time. “Artificial intelligence is indeed a turning point with enormous potential to advance medicine, accelerate scientific discoveries, expand opportunities and improve lives. Yet AI harbours risks, though the danger does not lie in the technology itself. “The danger lies in technology without accountability, in capabilities without oversight, in decision-making without transparency. The developers of these systems are now sounding the alarm.” Guterres proposed the creation of a multilateral “AI Risk Management Framework”.[52]
Western nations supported the negotiation of international rules, universal and interoperable technical standards, and independent oversight. French President Macron called for global AI regulation.[53] British Prime Minister Burham also advocated for international rules. “Fundamental decisions on this technology are too important to be left to chance. They must be taken by elected governments who serve working people.” Burham announced that the UK would prioritise AI during its G20 presidency in 2027.[54] By contrast, German Foreign Minister Wadephul remained very general in his remarks: “The companies that hold by far the largest share of all AI patents worldwide are based in just two major countries. This gives a very small group enormous influence. Such power entails responsibility, not only towards customers, shareholders or governments, but also towards humanity. And when power becomes global, responsibility must also become global.”[55]
US President Donald Trump firmly rejected plans for global regulation of AI and proposed replacing the term ‘AI’ with ‘superintelligence’. He argued that international regulations would stifle innovation.
Chinese Vice-President Han Zheng called for a “human-centred” framework to bridge the digital divide, whilst warning against the misuse of technological barriers as a tool of geopolitical protectionism.
Speakers from the “Global South” warned against “digital colonialism”. They noted that 90 per cent of the world’s population controls less than 1 per cent of global computing power, putting the Global South at risk of becoming entirely dependent on foreign infrastructure. Developing countries must become equal partners in shaping AI rules, rather than merely consumers.
On the eve of the 81st UN General Assembly, on 22 September 2026, 22 heads of state signed a “Call for Control of Frontier AI Models”.[56] The call was initiated by the Prime Minister of Norway, Jonas Gahr Store, and the President of Finland, Alexander Stubb. The call states: “AI must remain under human direction, oversight and control. It must be developed and used in line with international law.” The international community is called upon to begin discussions on the creation of a global AI organisation “to explore creating an international institution, able to set standards, enable verification, and convene states when capability thresholds are crossed”. The appeal has been signed by numerous heads of state and government from Europe, Australia and Canada, including Federal Chancellor Merz and European Commission President von der Leyen. The signatories also include the presidents of Kenya, Kazakhstan, the United Arab Emirates, South Africa and Singapore.

G20 Summit: AI Education Pact and Principles for New Technologies

The meeting of G20 digital ministers on 2 September 2026 in Chapel Hill, North Carolina, was effectively an AI education summit. The primary focus was on how the population can and must be prepared for the impending upheavals in the labour market. The ‘G20 Innovation Ministerial Statement’ set out six areas on which the G20 intends to focus: 1. Framework conditions that promote innovation; 2. Technology for opportunities and prosperity; 3. Developing a skilled technical workforce; 4. Guidelines on the protection of intellectual property in the field of AI; 5. Standards for AI; and 6. Industrial innovation and investment in supply chains.[57]

In a dedicated ‘AI Prosperity Compact’[58], the G20 has committed to investing in initiatives for the education and training of the workforce, in order to equip them with the knowledge and skills they need to succeed in an AI-driven labour market. “To ensure that the benefits of AI innovation are made available to the broader public, we will promote access to AI skills development for all our citizens, particularly those who have had less exposure to AI and STEM subjects to date.”[59]
In a further document, the ‘Carolina Principles for Emerging Technologies’, the G20 digital ministers agreed on 24 recommendations to promote research, commercialisation and application. The document emphasises innovation and the potential of AI, and recommends flexible regulations for managing risks. Section II.1 states: “Provide flexible regulatory sandboxes and experimental exemptions from select regulatory provisions where appropriate in supervised environments, whilst maintaining appropriate risk management as permitted under applicable legal frameworks, to reduce unnecessary barriers to scientific and technological research and accelerate scientific discovery”.[60] In his address to the UN Security Council on 21 September 2026, the Director of the White House Office of New Technologies, Michael Kraitsos, referred to these ‘Carolina Principles’ , which, in his view, constitute a sufficient framework for AI governance. Kraitsos rejected further discussions on AI governance.

European AI Strategy: Entry into force of the AI Act and new initiatives

In Europe, the AI Act came into force on 1 August 2026. At the same time, the ‘Digital Omnibus’ was adopted. In her ‘State of the Union’ address on 16 September 2026 in Strasbourg, the President of the European Commission, Ursula von der Leyen, discussed the next steps in the European AI strategy in detail.

The EU’s AI Act came into force on 1 January 2026. It is based on a so-called ‘risk-based approach’ and differentiates between risk categories according to the different rules to which they are subject. Since 1 August, for example, AI-generated content must now be labelled. On the same day, the ‘Digital Omnibus’ also came into force. This is intended to simplify the application of the regulations. In particular, small and medium-sized enterprises are to be relieved of bureaucratic reporting obligations. Some deadlines for high-risk applications have also been postponed until 2027 or 2028.[61]
On 16 September 2026, the President of the European Commission, Ursula von der Leyen, delivered her annual ‘State of the Union’ address in Strasbourg. She devoted a large part of her speech to the development of AI. She said she took the warnings from US AI companies about the risks of unchecked AI development seriously and proposed ‘shifting down a gear’. In November, von der Leyen plans to invite leading AI labs to discuss a joint strategy on model evaluation, verification, early warning and AI safety. By November 2026, the European Commission aims to present “pioneering initiatives” for the sectors of health, transport, agriculture and food, advanced manufacturing, as well as defence and space. Von der Leyen said: “We want AI. We want AI to create added value in a responsible manner. And we want it to do so at lower cost and with lower energy consumption. But above all, we want AI where humans remain in control. That is the essence of the European approach.” [62]

Alternative networks: WAICO, BRICS and the SCO form a counterweight

A rift is increasingly emerging in the global debate on AI. As a counterweight to the AI initiatives of the G7 countries (Hiroshima AI Process/HAIP), the OECD (Global Partnership for AI/GPAI) and the US (Pax Silica), alternative networks are taking shape, primarily under Chinese leadership. On 28 July 2026, a “World Artificial Intelligence Cooperation Organisation” (WAICO) was founded in Shanghai on the sidelines of a World AI Conference; 37 countries have joined so far. AI was also at the centre of discussions at the BRICS summits in New Delhi in September and the Shanghai Cooperation Organisation (SCO) summit in Bishkek.

On the fringes of the World AI Conference in Shanghai on 23 July 2026, 29 countries signed a treaty to establish a ‘World Artificial Intelligence Cooperation Organisation’ (WAICO).[63] Members include China, Russia, Brazil, Indonesia, Pakistan, South Africa, Serbia and Kazakhstan.

According to Article 3 of the treaty, WAICO’s purpose is to pursue a human-centred approach to the development of AI, to promote innovation, to strengthen international cooperation, to bridge the AI divide, to mitigate AI risks and to promote safe, ethical and trustworthy AI. In doing so, WAICO is to be guided by the principles of the UN Charter and work closely with the relevant UN bodies.
The highest governing body of WAICO is the WAICO-Council, in which all members are represented by one vote each. The aim is to adopt decisions by consensus. Should this not be possible, a two-thirds majority of the members present is sufficient. WAICO is managed by a secretariat based at the in Shanghai. It is to be headed by a Secretary-General and two Deputy Secretaries-General. The staff are intended to represent the breadth of the membership. However, as of the end of September 2026, the names of the leaders of the new WAICO had not yet been announced.
Commentators have described WAICO as a counterweight to the US’s ‘Pax Silica’. With ‘Pax Silica’, the US formed a network of like-minded states in December 2025 that support the US’s claim to develop the ‘gold standard’ for AI. WAICO primarily targets states from the Global South. By the end of September 2026, eight further states had acceded to the WAICO treaty, including Iran, Georgia and Vietnam. WAICO members used the UN General Assembly in September 2026 to promote membership. Kazakhstan is the only country that is a member of both WAICO and Pax Silica. Following the establishment of WAICO, Pax Silica had announced that dual membership was not possible. Consequently, India, which is a member of Pax Silica, refrained from joining WAICO, despite being a member of BRICS and the SCO.

The ten members of the BRICS network gathered in New Delhi on 12 September 2026 for their 20th summit. In the 45-page final declaration[64] the heads of state and government addressed issues including cyber security, the digital economy and artificial intelligence.

When it comes to cyber security, the BRICS states are focusing above all on strenghthening the UN. The creation of the new Global Mechanism on Cyber (G-MEC) and the UN Convention against Cybercrime (Hanoi Convention) are highlighted as significant steps forward. The commitment to an “open, safe, secure, stable, accessible, peaceful, and interoperable ICT environment whilst respecting the sovereignty and national security of each country” is reaffirmed. Support is given to the call, led primarily by Russia, for the drafting of an additional protocol to the Hanoi Convention, which is intended, amongst other things, to address the issue of illegal content on the Internet. The improved cooperation between CERTS and law enforcement agencies in combating cyber attacks “such as threats to security in the use of ICTs, malicious software, data security, and the misuse of technologies, including cybercrime, misinformation, hate speech, disinformation and deepfakes“ is acknowledged. Support is expressed for the BRICS Task Force on the security of submarine cables, which was established following the 2025 BRICS Summit in Rio de Janeiro. The call for new state regulation of Internet governance, advocated primarily by Russia, is not included in the declaration.
On the subject of the digital economy, progress has been noted in the New Industrial Revolution Programme (PartNIR), particularly with regard to closer cooperation between start-ups. The establishment of a BRICS Incubator Network is welcomed. However, no agreement has yet been reached on a BRICS Startup Innovation Fund.
On the subject of AI, reference is also made to the central role of the UN. The AI summit hosted by India in February 2026 as part of the Bletchley Process and the World AI Conference in Shanghai at the end of July 2026 are recognised as important milestones in involving the Global South in AI discussions. There is no mention of WAICO in the declaration. The ‘BRICS Leaders’ Statement on the Global Governance of Artificial Intelligence’[65] from 2025 is reaffirmed.

On 1 September 2026, the ten heads of state and government of the Shanghai Cooperation Organisation (SCO), including the Presidents of China, Xi Jiping; Russia, Vladimir Putin; and India, Naradem Modi, met in Bishkek for their annual summit. The SCO celebrated its 25th anniversary in Bishkek.

In the final document, the member states express their concern about the growing risks in the field of information security and categorically reject the militarisation of cyberspace and threats to the security of critical information infrastructure. They call for equal rights for all states in the regulation of the Internet and “consider that the provision of non- authorised satellite internet services within the territory of member states without the relevant licence and/or authorisation from their competent national authorities constitutes a breach of international telecommunications law.“[66]
The SCO members emphasise the key role of the United Nations in the field of international information security and call for the signing of the ‘Hanoi Convention’ against cybercrime. They reject the use of the Internet for terrorist, separatist and extremist purposes and will step up the coordination of their measures to combat new forms of cybercrime.
Also on the subject of AI, the SCO states are committed to close cooperation with the UN. There is no reference to WAICO in the declaration. India has not signed the WAICO treaty. The signing of an SCO memorandum on cooperation in the field of artificial intelligence was announced. In the memorandum, the SCO member states have agreed on a strategic framework for establishing regional data processing centres and creating high-performance computing capabilities.[67]

III. Cybersecurity

AI is also playing an increasingly significant role in the field of cyber security. This applies in particular to the military dimension. As the Ukraine crisis in particular demonstrates, AI has changed the nature of warfare. However, AI is also playing an increasingly significant role in the fight against cybercrime. When Anthropic’s Claude was able to identify hundreds of vulnerabilities in functioning software programmes, alarm bells rang. If such AI were to fall into the hands of cybercriminals, no application on the Internet would be safe. Leading AI developers are issuing increasingly stark warnings about autonomous AI (whether with good or bad intentions) infiltrating external networks. Bill Gates, Sam Altman, Mario Amodei and others cannot rule out the possibility that autonomous AI might itself become a weapons developer and create chemical or biological weapons prohibited under international law. These threats are also increasingly shaping international cybersecurity negotiations. In the third quarter of 2026, the key events were the inaugural meeting of the new ‘Global Mechanism for Cyber Security’ (G-MEC), held from 20 to 24 July 2026 in New York, and the meeting of the Working Group on Autonomous Weapon Systems (GGE LAWS), held from 1 to 5 September 2026 in Geneva.

Global Mechanism for Cyber Security (G-MEC): Focus on capacity building

The new ‘Global Mechanism for Cyber Security’ (G-MEC) held its first meeting in New York at the end of July 2026. The G-MEC replaces the Open-Ended Working Group (OEWG), which has been addressing the issue of cyber security within the UN over the past five years. The G-MEC is now a permanent body within the UN and the central forum for cybersecurity negotiations. Its mandate largely follows that of the OEWG. It focuses on the analysis of threat scenarios and the application of international law in cyberspace, as well as on confidence-building and capacity-building measures. One concrete outcome of the OEWG was the establishment of a ‘Global Intergovernmental Points of Contact Directory (POC Directory)’[68] as a confidence-building measure. The POC Directory will continue under the G-MEC. The new Chair of the G-MAC is Egriselda López, El Salvador’s Ambassador to the UN. She is also Co-Facilitator of the ‘Global Dialogue on AI Governance’.

Despite geopolitical tensions, the July 2026 meeting of the G-MEC took place in a constructive and results-oriented atmosphere. Capacity-building measures were the focus of the meeting. Following a proposal by India, it was decided to establish a ‘Global ICT Security Cooperation and Capacity Building Portal’[69], which will ensure that all states have access to the necessary information on cyber security. Capacity-building measures should be “demand-driven, nationally owned, and sustainable”. Care should be taken to ensure that women are specifically included in relevant programmes. It was also recommended to collaborate with regional organisations to avoid duplication. The second meeting of the G-MEC is scheduled for 7-11 December 2026 in New York.
The first meetings of the two ‘Dedicated Thematic Groups’ (DTGs) are also scheduled to take place alongside this December meeting. The DTGs are, in effect, sub-organisations of the G-MEC. They focus on the protection of critical infrastructure against cyber attacks (DTG-1) and on capacity-building measures (DTG-2). However, a dispute has arisen over the chairpersons during the preparations for the DTG meetings. The idea that the G-MEC Chair should nominate the two chairpersons at their own discretion was contested by some countries, including Russia. They demanded that the chairpersons be appointed by consensus among all states. The involvement of non-state stakeholders also remains unclear. Initially, it had appeared that the procedure for their participation in the DTGs would be more flexible. However, Russia has now demanded that the arrangement reached at the end of the OEWG – whereby, in addition to NGOs accredited to ECOSOC, other non-state actors from business, academia and civil society may participate without the right to speak under certain conditions, provided they are not subject to a reasoned veto – should also apply to the DTGs.[70]

GGE LAWS: Consensus report on autonomous weapon systems adopted

The working group on autonomous weapon systems (GGE LAWS) held what is, for the time being, its final session in Geneva from 1 to 5 September 2026. The aim of the session was to reach agreement on elements of a future legally binding protocol under international law on autonomous weapon systems. Following contentious negotiations, a heavily watered-down final report[71] was adopted by consensus - a result already regarded as a success given the prevailing geopolitical tensions. The US, Russia, Turkey and Israel, in particular, had repeatedly insisted on recommendations that were as non-binding as possible. On the other hand, 78 states had spoken out in favour of an instrument under international law.

On the eve of the conference, UN Secretary-General Antonio Guterres and the President of the International Red Cross, Mirjana Spoljaric Egger, had appealed to the negotiators to reach a legally binding instrument as soon as possible that prohibits autonomous weapons without human control. They called for “political courage” and warned that “a failure to act quickly and decisively will cost lives. It will leave future generations to inherit a world in which the absence of clear and specific prohibitions and restrictions on these weapons opens the door to practices that would erode the protections international humanitarian law is meant to guarantee.”[72]
The main point of contention was the proposed definition of autonomous weapon systems. The report now proposes the following definition: “Within the scope of the application of the CCW, a lethal autonomous weapon system can be characterised as a combination of one or more weapons and technological components operating together, which is capable of identifying, selecting and engaging a target in a way that results in loss of life, without intervention by a human being operating the system.” (Within the scope of the CCW, a lethal autonomous weapon system can be characterised as a combination of one or more weapons and technological components that operate together and are capable of identifying, selecting and engaging a target in such a way that this results in loss of life without human intervention.) In previous sessions, the GGE LAWS had opted for a so-called two-tier approach. Accordingly, autonomous weapons without human control should be banned. Autonomous weapons under human control should be subject to the Geneva Conventions on international humanitarian law. This is now confirmed in the final report, which consistently states that any use of autonomous weapons in armed conflicts is subject to the provisions of the Geneva Conventions on international humanitarian law.
Critics such as the NGO “Stopp Killer Robots” (SKR) criticised the fact that, during the final negotiations, references in the final report to constructive concepts and safeguards – such as those relating to “predictability, reliability, explainability and traceability” – had disappeared[73]. Nevertheless, the report has been well received and is regarded, even by SKR, as a sound basis for commencing negotiations.
The report will now be submitted to the Conference of the States Parties to the Convention on Certain Conventional Weapons (CCW), which is due to meet in Geneva in November 2026[74]. Under the umbrella of this 1980 Convention, legally binding protocols on laser weapons, cluster munitions and landmines have already been drawn up. The conference must now decide whether to commence formal negotiations on autonomous weapon systems.