As the UN prepares its first Global Dialogue on AI Governance, communities across the Global Majority are clear: they need more than a seat at the table. They need the table itself to change. Through our contribution in the written submission process, we proposed that the Co-Chairs’ Summary documents genuine convergence and divergence by stakeholder type and region; that the Dialogue rejects the assumption that Global North frameworks define the baseline for AI governance; that the Dialogue commits to building contextual AI safety infrastructure for the Global Majority; and that the Dialogue agrees to a concrete pilot, such as a regional AI audit unit, with a timeline.
Across South Asia, West Africa, Latin America, and beyond, AI systems are already shaping who receives welfare, who gets credit, who is surveilled, and whose language a system understands. Yet the rules governing these systems are being written almost entirely by a small number of Global North governments and corporations. The problem is structural. Most AI models are built on a small amount of Western data, optimized for specific Western contexts, and deployed in Global Majority countries with no local oversight. There is no regional body in West Africa, South Asia, or the Andean region that can independently check whether a AI system outside of the region is safe before it is used to decide someone’s welfare claim or visa application.
Women and girls are disproportionately harmed when systems fail in local languages, when biometric tools exclude them, or when workers face wage discrimination and harassment with no recourse. Indigenous knowledge is scraped into training datasets without consent. Gig workers face algorithmic dismissal with no right of appeal.
Meanwhile, frameworks from the EU or the US are treated as the global standard, while a law from Brazil, a regulation from Indonesia, or a policy from Kenya is dismissed as a “regional adaptation.”
Our submission raises four further cross-cutting issues requiring explicit attention: power asymmetries and data justice, the environmental costs of AI infrastructure, labor and algorithmic management, and AI in conflict and surveillance contexts. We center our contributions by acknowledging the structural dependency of Global Majority countries resulting in disproportionate harms faced by women, girls, and gender-diverse people, and the exclusion of data workers, gig workers, Indigenous peoples, and low-resource language communities from governance design. We urge the Co-Chairs to underpin the Dialogue and Summary in international human rights frameworks as this provides the most credible and universal baseline for global cooperation.
We call on the Co-Chairs with five specific asks:
- Name the disagreements. The Co-Chairs’ Summary must record where genuine agreement exists and where it does not — by region and stakeholder type. Vague language that papers over conflict only serves the status quo.
- Treat all frameworks as equal. Governance contributions from the Global Majority must be discussed as legitimate approaches in their own right, not as departures from a standard set elsewhere.
- Fund regional AI audit capacity. Dedicated investment in multilingual evaluations, locally grounded benchmarks, and independent audits conducted by regional institutions and civil society is non-negotiable. AI safety claims must be backed by transparent disclosure of test conditions, failure rates by language and gender, and known limitations.
- Agree on a concrete pilot with a deadline. Establish one regional AI audit unit, hosted by an existing institution in the Global Majority, with a mandate to audit deployed public sector AI systems before the second Dialogue. Gender-disaggregated harm reporting must be a core requirement.
- Ensure structural parity in facilitating multistakeholder participation by establishing a transparent, inclusive selection criteria that distinguish grassroots community organizations from well-resourced intermediaries who may claim to represent them.
Global Majority advocates have long argued that the wheel does not need to be reinvented. Existing frameworks, such as the Internet Governance Forum coupled with the regional and national IGFs, and the NETmundial+10 process have spent years building the infrastructure for inclusive, multistakeholder participation in digital governance. The Co-Chairs should treat them as foundations in multistakeholder parity and transparency to facilitate the Dialogue. The IGFs national and regional network means that governance conversations do not have to start from scratch each time a new global process is launched. Communities in every region already participate in regional IGFs that are conducted in local languages and are grounded in locally specific challenges. Feeding these into a global AI Dialogue requires deliberate design. The Dialogue’s design must further ensure structural parity, shifting away from the the norm in most UN processes, where government delegations hold formal decision-making authority and civil society is admitted as observers. It is therefore critical to establish a transparent, inclusive selection criteria that distinguish grassroots community organizations from well-resourced intermediaries who may claim to represent them.
The NETmundial+10 process set a standard that the AI Dialogue should adopt explicitly: every synthesis document should explain how community inputs shaped the final output, including where inputs were not adopted and why. This is not a small procedural matter. For Global Majority civil society organizations that invest significant resources in preparing submissions, thereby a summary that simply absorbs all views into vague consensus language erases the vast diversity and plurality of Global Majority realities. Naming disagreements, by region and stakeholder type, is what gives the process legitimacy and gives communities a record they can use in national and regional advocacy afterward.
Our contribution further warns that one-off summits produce one-off commitments. The IGFs intersessional working groups, dynamic coalitions, and year-round secretariat work mean that conversations continue, commitments are tracked, and communities do not have to rebuild relationships from zero each year. The Dialogue needs to adopt a similar anchoring in an intersessional accountability mechanism, potentially within the IGF secretariat (rather than creating yet another parallel body) that could use existing capacity into AI governance.
We recognize even the aforementioned existing processes may not sufficiently accommodate diverse grassroots voices. We call on the UN and the Co-Chairs to ensure that participation in the Dialogue moves beyond symbolic presence. Affected communities, such as gig workers, indigenous knowledge holders, and women excluded from biometric systems, can give direct testimony to policymakers and decision-makers who must respond publicly. The Dialogue should form multi-stakeholder working groups that tackle real design gaps and implementation problems with deadlines, not deliver prepared statements. Civil society from the Global Majority needs resourcing to meaningfully participate, including bringing grassroots voices early on and ensuring wide-ranging participating across countries, communities, languages and political contexts, as well as funded fellowships and visa support. In addition, the Dialogue’s intersessional framework must include named indicators, public reporting, and a formal mechanism for civil society to flag non-delivery as accountability mechanisms for states and corporations if they do not follow through commitments.
In addition to this standalone submission, we are pleased to have contributed to the joint submissions coordinated by Global South Alliance (GSA), Global Digital Justice Forum (GDJF), South Asians for Digital Rights (SADR), and Multistakeholder Approaches to Participation in AI Governance (MAP-AI), to amplify shared priorities. We further support the submission from Global Digital Rights Coalition (GDRC), another coalition we are proud to be a part of.
You can read our full submission here.
The UN’s First AI Governance Dialogue Must Prioritize the People Most Affected by AI, Not Just Those Who Build It
As the UN prepares its first Global Dialogue on AI Governance, communities across the Global Majority are clear: they need more than a seat at the table. They need the table itself to change. Through our contribution in the written submission process, we proposed that the Co-Chairs’ Summary documents genuine convergence and divergence by stakeholder type and region; that the Dialogue rejects the assumption that Global North frameworks define the baseline for AI governance; that the Dialogue commits to building contextual AI safety infrastructure for the Global Majority; and that the Dialogue agrees to a concrete pilot, such as a regional AI audit unit, with a timeline.
Across South Asia, West Africa, Latin America, and beyond, AI systems are already shaping who receives welfare, who gets credit, who is surveilled, and whose language a system understands. Yet the rules governing these systems are being written almost entirely by a small number of Global North governments and corporations. The problem is structural. Most AI models are built on a small amount of Western data, optimized for specific Western contexts, and deployed in Global Majority countries with no local oversight. There is no regional body in West Africa, South Asia, or the Andean region that can independently check whether a AI system outside of the region is safe before it is used to decide someone’s welfare claim or visa application.
Women and girls are disproportionately harmed when systems fail in local languages, when biometric tools exclude them, or when workers face wage discrimination and harassment with no recourse. Indigenous knowledge is scraped into training datasets without consent. Gig workers face algorithmic dismissal with no right of appeal.
Meanwhile, frameworks from the EU or the US are treated as the global standard, while a law from Brazil, a regulation from Indonesia, or a policy from Kenya is dismissed as a “regional adaptation.”
Our submission raises four further cross-cutting issues requiring explicit attention: power asymmetries and data justice, the environmental costs of AI infrastructure, labor and algorithmic management, and AI in conflict and surveillance contexts. We center our contributions by acknowledging the structural dependency of Global Majority countries resulting in disproportionate harms faced by women, girls, and gender-diverse people, and the exclusion of data workers, gig workers, Indigenous peoples, and low-resource language communities from governance design. We urge the Co-Chairs to underpin the Dialogue and Summary in international human rights frameworks as this provides the most credible and universal baseline for global cooperation.
We call on the Co-Chairs with five specific asks:
Global Majority advocates have long argued that the wheel does not need to be reinvented. Existing frameworks, such as the Internet Governance Forum coupled with the regional and national IGFs, and the NETmundial+10 process have spent years building the infrastructure for inclusive, multistakeholder participation in digital governance. The Co-Chairs should treat them as foundations in multistakeholder parity and transparency to facilitate the Dialogue. The IGFs national and regional network means that governance conversations do not have to start from scratch each time a new global process is launched. Communities in every region already participate in regional IGFs that are conducted in local languages and are grounded in locally specific challenges. Feeding these into a global AI Dialogue requires deliberate design. The Dialogue’s design must further ensure structural parity, shifting away from the the norm in most UN processes, where government delegations hold formal decision-making authority and civil society is admitted as observers. It is therefore critical to establish a transparent, inclusive selection criteria that distinguish grassroots community organizations from well-resourced intermediaries who may claim to represent them.
The NETmundial+10 process set a standard that the AI Dialogue should adopt explicitly: every synthesis document should explain how community inputs shaped the final output, including where inputs were not adopted and why. This is not a small procedural matter. For Global Majority civil society organizations that invest significant resources in preparing submissions, thereby a summary that simply absorbs all views into vague consensus language erases the vast diversity and plurality of Global Majority realities. Naming disagreements, by region and stakeholder type, is what gives the process legitimacy and gives communities a record they can use in national and regional advocacy afterward.
Our contribution further warns that one-off summits produce one-off commitments. The IGFs intersessional working groups, dynamic coalitions, and year-round secretariat work mean that conversations continue, commitments are tracked, and communities do not have to rebuild relationships from zero each year. The Dialogue needs to adopt a similar anchoring in an intersessional accountability mechanism, potentially within the IGF secretariat (rather than creating yet another parallel body) that could use existing capacity into AI governance.
We recognize even the aforementioned existing processes may not sufficiently accommodate diverse grassroots voices. We call on the UN and the Co-Chairs to ensure that participation in the Dialogue moves beyond symbolic presence. Affected communities, such as gig workers, indigenous knowledge holders, and women excluded from biometric systems, can give direct testimony to policymakers and decision-makers who must respond publicly. The Dialogue should form multi-stakeholder working groups that tackle real design gaps and implementation problems with deadlines, not deliver prepared statements. Civil society from the Global Majority needs resourcing to meaningfully participate, including bringing grassroots voices early on and ensuring wide-ranging participating across countries, communities, languages and political contexts, as well as funded fellowships and visa support. In addition, the Dialogue’s intersessional framework must include named indicators, public reporting, and a formal mechanism for civil society to flag non-delivery as accountability mechanisms for states and corporations if they do not follow through commitments.
In addition to this standalone submission, we are pleased to have contributed to the joint submissions coordinated by Global South Alliance (GSA), Global Digital Justice Forum (GDJF), South Asians for Digital Rights (SADR), and Multistakeholder Approaches to Participation in AI Governance (MAP-AI), to amplify shared priorities. We further support the submission from Global Digital Rights Coalition (GDRC), another coalition we are proud to be a part of.
You can read our full submission here.
Related Resources
Consultation
Announcements
Research