Rwanda President Paul Kagame, Salesforce Chair and CEO Marc Benioff, and International Telecommunication Union Secretary-General Doreen Bogdan-Martin have launched the AI for Good Global Commission, an ITU-led, UN-associated body intended to bring governments, technology companies and international institutions into a single forum for AI deployment and governance. Announced July 2, the 44-member commission will hold its inaugural meeting in Geneva on July 8 during the AI for Good Global Summit. [1]
The initiative matters less as a new AI regulator than as an attempt to coordinate the infrastructure, standards, investment and public-sector capacity behind AI adoption. Its stated ambition spans trust and safety, access to compute and connectivity, skills, and applications in health, education, food security and disaster response. But as the summit opens on July 7, the commission has not yet published binding commitments, a funding plan, technical standards or an enforcement framework.
A public-private commission with heavyweight backing
Kagame and Benioff will serve as co-chairs, with Bogdan-Martin as vice-chair. The membership combines heads of state, senior ministers, international-agency leaders and executives from companies that sit across much of the AI supply chain.
Among the listed founding commissioners are Amazon CEO Andy Jassy, NVIDIA founder and CEO Jensen Huang, Microsoft Vice Chair and President Brad Smith, Google and Alphabet research and technology leader James Manyika, Anthropic co-founder Jack Clark, Cohere CEO Aidan Gomez, Sakana AI co-founder Ren Ito, Qualcomm CEO Cristiano Amon, and Reliance Industries Chairman Mukesh Ambani. Telecom, consulting and enterprise-technology leaders including Vodafone’s Margherita Della Valle, MTN’s Ralph Mupita, Orange’s Christel Heydemann, Accenture’s Julie Sweet and ZTE’s Xu Ziyang are also represented. [1]
The public and multilateral side includes the presidents of Estonia and Iceland; officials from Kazakhstan, Namibia, Nigeria, Singapore and Togo; and organizations including the African Union, UNDP, UNESCO, the World Intellectual Property Organization, the World Trade Organization and GSMA.
The structure follows the ITU and UNESCO Broadband Commission for Sustainable Development, which uses a public-private model to address connectivity and digital inclusion. The newer commission applies that approach to AI at a point when the technology’s global impact depends increasingly on assets beyond the models themselves: data centers, power, network coverage, cloud access, skilled workers and public institutions able to procure and oversee systems. [1]
Its agenda is as much about access as ethics
“AI for Good” language often evokes high-level principles, but the commission’s announced scope is rooted in the practical conditions needed to deploy AI. The ITU says 2.2 billion people remain offline, or roughly one-quarter of the world’s population. For countries without extensive domestic cloud infrastructure or frontier-model developers, that connectivity gap is also an AI gap. [1]
The resulting agenda includes broadband and mobile access, cloud and data-center capacity, affordable compute and energy, local-language data and models, digital public infrastructure, and workforce training. These questions determine whether countries can adapt and deploy AI to local needs or are limited to consuming services designed and hosted elsewhere.
That focus could make the forum particularly relevant to developing economies. A health-service AI system, for example, requires more than a capable model: it may depend on reliable connectivity, protected health data, local-language interfaces, trained clinical staff, procurement safeguards and a way to evaluate performance in local settings. Similar constraints apply to education, disaster response and food-security projects.
The commission also arrives as AI policy is expanding from oversight of model developers into oversight of the deployment stack. Chips, electricity, cooling, telecom networks, enterprise software, public procurement and autonomous systems are becoming governance issues because they shape who can build, use and benefit from AI.
Trust and safety without a published rulebook
Benioff has described the initial work as strengthening AI infrastructure, expanding public-interest applications and ensuring trust and safety. The wider AI for Good Summit program points to a technically broad set of concerns: agentic-AI security, model testing and benchmarking, misinformation and deepfakes, energy demand, edge AI and TinyML, robotics, brain-computer interfaces, quantum applications and space-based computing. [2]
Agentic systems deserve particular attention because they can plan and carry out multistep actions rather than simply generate a response. Their risks can involve tool access, permissions, unintended actions and the challenge of tracing responsibility when systems act through other software or services. Effective governance in that area could require common evaluation methods, controlled testing, incident reporting and clear accountability across developers and deployers.
None of those mechanisms has been announced for the commission. As of July 7, there is no public model-evaluation protocol, audit requirement, shared incident-reporting system, model-access policy or technical standard that members have agreed to follow. There is also no indication that the body will directly supervise AI companies or issue rules with the force of law.
That limitation is central to understanding the initiative. The commission is separate from the UN General Assembly-convened Global Dialogue on AI Governance, held in Geneva on July 6 and 7. Its role appears closer to coordination and implementation than treaty-making or regulation. [1]
Why coordination is difficult
The commission enters an increasingly fragmented governance environment. National laws, voluntary corporate commitments, sector-specific rules and technical standards are developing at different speeds and with different priorities. The UN University has warned that this patchwork can create governance arbitrage, in which firms or deployments shift toward jurisdictions with more favorable or weaker requirements. [3]
Stanford’s AI Index has likewise described a widening gap between rapidly advancing capabilities and investment on one side, and governance, evaluation and transparency mechanisms on the other. [4] A forum that includes cloud providers, chipmakers, telecom companies, model developers, governments and international organizations could help narrow practical gaps where no individual participant controls the entire system.
For companies, participation provides an opportunity to shape emerging international norms, procurement priorities and standards. For governments, particularly those seeking AI capacity, it creates direct access to the firms controlling crucial infrastructure and to potential partners in training and deployment. The 2026 AI for Good Summit itself has drawn substantial industry backing, while the prior year’s combined AI for Good and WSIS events attracted more than 11,000 participants from 169 countries. [5]
Those same links create an accountability challenge. Many commissioners lead companies with direct commercial interests in cloud services, chips, telecom networks, consulting, enterprise software or AI deployment. The launch materials do not say how conflicts of interest will be disclosed or managed, how voluntary pledges would be measured, or whether progress will be independently audited.
The representation test
The commission emphasizes inclusion and developing-country participation, yet its founding membership is heavily weighted toward political leaders, corporate executives and intergovernmental organizations. The published list does not identify independent civil-society organizations or academic researchers as founding commissioners. [1]
That absence could become consequential in decisions involving labor displacement, discrimination, privacy, human rights, content moderation and harms experienced by communities without ownership of AI infrastructure. Technical capacity and investment are necessary for broad-based AI access, but they do not by themselves establish legitimacy or ensure that affected groups have a meaningful role in setting priorities.
UN work on AI governance has stressed both the territorial fragmentation of current rules and the need for interoperability among binding law, technical standards and voluntary measures. It has also emphasized inclusive participation. [6] The commission will be judged in part on whether it can convert its unusually senior membership into concrete, transparent work without allowing access to become a proxy for representation.
Its first real test is still ahead. The July 8 inaugural meeting may clarify priorities, deliverables and governance arrangements, but no outcomes can yet be attributed to it. For now, the AI for Good Global Commission is a significant convening effort with broad ambitions: to align the institutions that build AI, the countries that regulate and deploy it, and the infrastructure needed to make its benefits more widely available. Whether it becomes a vehicle for shared implementation or mainly another venue for high-level dialogue will depend on the commitments it makes next.
Editor’s Take
I see the commission’s strongest potential not in writing another set of AI principles, but in making deployment less extractive. Reliable networks, affordable compute, local-language data, capable procurement teams and electricity are the unglamorous dependencies that determine whether an AI pilot becomes a useful public service or a short-lived vendor demonstration. Getting chipmakers, cloud platforms, telecom operators and governments in the same room can remove real bottlenecks—if the group publishes specific commitments with owners, timelines and measurable outcomes.
The near-term test is whether July’s launch produces operating machinery rather than summit language: interoperable evaluation practices, incident-reporting pathways, public-sector procurement templates, financing for connectivity and compute, and transparent reporting on who benefits. The membership is commercially powerful, which can accelerate execution, but it also makes conflict rules and independent oversight non-negotiable. Without civil-society and research voices at the table, “access” risks becoming a sales channel for infrastructure rather than a durable model of local control and accountability.
References
- International Telecommunication Union – https://www.itu.int/en/mediacentre/Pages/PR-2026-07-02-AI-for-Good-Global-Commission.aspx
- Axios – https://www.axios.com/2026/07/01/un-ai-commission-ceos-world-leaders
- UN University – https://unu.edu/article/ai-governance-arbitrage
- Stanford Institute for Human-Centered Artificial Intelligence, AI Index – https://hai.stanford.edu/ai-index
- International Telecommunication Union – https://www.itu.int/en/mediacentre/Pages/PR-2026-03-25-AI-for-Good-Global-Summit.aspx
- United Nations, Toward Interoperability and Inclusive Participation in AI Governance – https://www.un.org/digital-emerging-technologies/sites/www.un.org.digital-emerging-technologies/files/Toward_Interoperability_and_Inclusive_Participation_in_AI_Governance_White_Paper.pdf
,_Geneva,_Switzerland.jpg){kind=link}