The India AI Impact Summit, positioned to empower Global South voices, has been met with hesitation as experts questioned whether the New Delhi convening and Declaration can actually alter the trajectory of international AI governance for good. The Summit’s Sanskrit slogan ‘Sarvajan Hitaya, Sarvajan Sukhaya,’ translating to ‘Welfare for all, Happiness for all’, captured the host country’s focus on human-centric, inclusive AI growth, which directly challenged previous iterations’ narrow agendas, extending the outcomes and expectations to focus on AI as a development pathway to reduce digital divides, and not just a tool for safety and risk mitigation. Similarly, expanded participation from more than 100 countries immediately promised more inclusive representation.
Yet, even with 88 countries signing off on the final New Delhi Declaration, the question remains: What does this mean for Global South-led AI governance?
The road to New Delhi 2026
The seat of the global AI Summit has long been held by Global North actors, whose rotating chairs have included the United Kingdom (UK), Seoul and Paris. From the founding Bletchley Park event in the UK in November 2023 to the latest convening in New Delhi in 2026, Summit priorities have shifted from narrow AI safety strategies to a wider focus on inclusion, equity, and digitally-led development. With India framing its aspirations around three ‘chakras’, People, Planet and Progress, the agenda was certainly more public-interest facing than previous models.
The AI Safety Summit at Bletchley Park, 2023:
- The Bletchley Declaration was an international agreement among 29 governments that collectively committed to developing AI in “a manner that is safe, in such a way as to be human-centric, trustworthy and responsible.”
- Signatories were Global-North dominated, including the UK, United States (US) and China, as well as France, Germany, Italy, Spain and the European Union (EU).
- The Chair’s Statement on AI ‘Safety Testing’ was signed by 10 countries, including the UK, US, and major European member states, as well as leading technology companies.
- The UK government announced the formation of a new AI Safety Institute (AISI), which mirrored the Summit’s prioritisation of AI safety as a top AI policy concern.
The AI Seoul Summit, 2024:
- Framed as a ‘mini Summit’, the Seoul agenda adopted a notably broader framing, adding ‘innovation’ and ‘inclusivity’ to pre-existing commitments to AI safety.
- While the event saw reduced coverage and participation, it led to increased commitments to develop AI safety institutes from Japan, South Korea, and Canada, thereby increasing global government capacity in AI Safety.
- Korean and UK organisers secured a Statement of Intent Toward International Cooperation on AI Safety Science, signed by 10 countries and the EU, to cooperate as an international network of AI Safety Institutes. These countries were still largely Global North actors (Australia, Canada, France, Germany, Italy, Japan, the Republic of Korea, Singapore, the UK, and the US).
Paris AI Action Summit, 2025:
- Further shifting emphasis away from AI safety, the Summit cast a new focus on innovation, economic opportunity and international cooperation.
- Unlike previous declarations, the final Declaration recognised the need to promote AI accessibility to reduce digital divides. Other focal points included the need for positive labour impacts and the opportunity for AI-enabled innovation without market concentration.
- Sixty-one countries signed the Declaration; however, both the UK and the US rejected the Declaration, and in general, it was considered to have notably fewer concrete actions and outcomes.
- A USD 400 million investment was put towards a Current AI initiative for public interest AI, and the development of The Coalition for Environmentally Sustainable AI, which included 91 partners committed to addressing AI’s planetary impacts. These partners comprised 37 tech companies, 11 governments (including Chile, Denmark, Finland, France, Germany, India, Kenya, South Korea, Morocco, Norway, and the UK), and key international organisations such as the United Nations Development Programme, the International Energy Agency, and the Global Green Growth Institute.
The New Delhi 2026 AI Impact Summit
- Eighty-eight signatory nations committed to formally adopting frameworks for interoperable, sovereign AI and cross-border digital public infrastructure (DPI). Of these signatories, 75% were from the Global South, while 35% were from the Global North.
- Trends in narratives from key speakers showed the dominance of themes such as ‘sovereignty’ and ‘AI for all’ in discussions, reflecting aims to enhance access to AI services as well as concerns about data sovereignty in the data governance frameworks that shape AI training, modelling, and deployment.
- USD 250 billion dollars in capital was dedicated to scaling infrastructure and localised research and development programming.
Guided by the principle of ‘Sarvajan Hitaya, Sarvajan Sukhaya’, or ‘Welfare for all, Happiness for all’, the Declaration underscores that the benefits of AI must be equitably shared across humanity. The Declaration highlights the role of AI in driving economic transformation; the importance of open-source and accessible AI ecosystems; the need for energy-efficient AI infrastructure; and the call to expand AI’s role in science, governance, and public service delivery.
The commitment? Advancing shared global priorities in AI governance: an approach that looks to strengthen international cooperation and multistakeholder engagement on the one hand, and respect national sovereignty on the other. This is proposed through a non-binding framework that promotes long-term international partnerships through a series of voluntary, collaborative global initiatives, which position AI as a key driver of economic growth.
The Summit announced a series of voluntary, collaborative global initiatives, including:
Charter for the Democratic Diffusion of AI | Promotes affordable access to foundational AI resources; Supports locally relevant innovation ecosystems. |
| Global AI Impact Commons | A platform to scale and replicate AI use cases globally; Enables cross-country collaboration for development impact. |
| Trusted AI Commons | Repository of tools, benchmarks, and best practices; Supports the development of secure and trustworthy AI systems. |
| International Network of AI for Science Institutions | Facilitates global scientific collaboration; Enhances AI-driven research capabilities. |
| AI for Social Empowerment Platform | Enables knowledge exchange and scalable solutions; Focuses on equitable AI adoption. |
| AI Workforce Development Playbook & Reskilling Principles | Supports AI skilling, reskilling, and literacy; Prepares nations for an AI-driven economy. |
| Guiding Principles on Resilient & Efficient AI | Focus on energy-efficient AI systems; Supported by a Playbook on AI Infrastructure Resilience. |
However, while the Declaration achieved wider buy-in from a more representative group of signatories, its governance commitments were minimal and operationally thin, with investment outshining any binding agreements. Additionally, despite a human-centric selection of ‘chakras’, these guiding principles were not practically interrogated in the final Declaration, with little resolve on how to manage gaps in infrastructure, compute capacity or capital funding when attempting to ‘democratise’ AI, for example. And unlike the establishment of the Coalition for Environmentally Sustainable AI in Paris, or the commitments to AI safety Institutes in Seoul and the UK, India’s outcomes saw no new multilateral coalitions formed among middle powers – a missed opportunity to concretise the multilateral agenda at the Summit.
Global South-led AI governance: Bringing justice, equity, and evidence-based research into focus
Remaining critical of what constitutes strategic versus symbolic participation, our contribution to the Summit expanded beyond Bharat Mandapam. Bringing forth our evidence base from the flagship Africa Just AI project, we aligned with multistakeholder partners to discuss what global AI governance could truly look like when grounded in participatory processes and the lived realities of human citizens across continents. Connecting with partners from the ‘Multistakeholder Approaches to Participation in AI Governance (MAP-AI)’ initiative, the Global South Alliance and the Global Digital Justice Forum, among others, our engagements confirmed that coordinated action is still needed to operationalise the commitments made between 16 and 20 February.
Furthermore, Just AI governance approaches that bridge the gap between the Summit’s value-driven approach and the need for more technical, rooted policy and governance solutions are still necessary. Below is the agenda that we will be taking forward:
- We are advocating for the virtuous production of data. This involves creating frameworks where local language broadcasters and community archives develop a sovereign data exchange capability for sustained access to high-quality, culturally nuanced data needed for local LLMs.
- Global AI safety discussions often focus on frontier risks, while ignoring the immediate harms prevalent in our regions—such as AI-enabled targeted hate, discrimination, and the erosion of democratic fundamentals in elections. As part of our Just AI Framework of Inquiry, we are focusing on a safety agenda that prioritises lifecycle accountability (from design to post-deployment monitoring) for the immediate harms prevalent in the Global South.
- There is a structural opportunity to leapfrog legacy industrial models, but it requires a “compass” that points toward justice rather than just productivity. We are positioning Just AI as that compass. Our 2026 Africa Just AI Conference brings together edge thinkers from across Africa in March, and we will leverage this convening to further develop Just AI’s anticipatory evaluation methodology to help governments assess if AI deployment narrows or widens the ‘sovereignty divide’.
- We are researching the blueprint for AI-enabled economic growth with Africa at the intersection of digital trade and data commons, with a commitment to data sovereignty as the bedrock. The goal is an interoperable data infrastructure and data resources that allow African Small and Medium Enterprises to trade seamlessly.
- We will be watching for catalytic impact. With over USD 200 billion in AI investment pledged at the Summit, our role must be to study impacts and assess whether these capital flows serve the public interest and local infrastructure, not just extractive frontier models.
- In 2026, the governance of environmental data has shifted from voluntary corporate reporting to a high-stakes arena of algorithmic accountability. As we champion the ‘for planet’ pillar of our work, we are moving beyond simple data access to a more radical concept: environmental data agency. Such agency depends on active transparency, including the reporting of environmental data in open, interoperable formats — especially those in extractive industries — to disclose high-resolution ecological data. We are calling for dynamic disclosure or real-time, machine-readable API access to environmental sensors (water quality, carbon output, soil health) as a condition of their digital license. Through better and more just AI governance, we are interested in how we would mitigate data washing, where centralised authorities aggregate (and thus obscure) localised ecological damage.
Towards Just AI futures
Governance must follow the Summit, because the “Delhi Consensus” only becomes real when it is codified. For Global South actors, the Summit must be followed with action, in solidarity with inclusive, equitable and just AI futures for the Global South. Here at Research ICT Africa, we believe the Summit has offered a valuable window into the transition of AI governance priorities and their remaining gaps. Our experience validates the need for our Just AI work, informed by our Framework of Inquiry, to cease the continuation and intensification of AI injustice.
