Reflections from MAP-AI: From Principles to Action in Global AI Governance

Over the past two days, Global Partners Digital (GPD) had the opportunity to participate in two multistakeholder dialogues convened under the Multistakeholder Approaches to Participation in AI Governance (MAP-AI) initiative, led by the Global Network Initiative (GNI) and the Centre for Communication Governance (CCG) at the National Law University Delhi. These convenings were designed to strengthen civil society engagement and enhance multistakeholder participation at the India AI Impact Summit.

MAP-AI’s broader objective is both timely and necessary: to foster meaningful, effective engagement across the growing number of AI governance processes and initiatives, while ensuring that underrepresented voices are not only present, but influential.

As the discussions conclude, we want to share a few reflections.

Revisiting the AI Governance Landscape

On Monday 16th, we contributed to a session titled “Decoding AI Jargon and Mapping Key Themes, Processes, and Initiatives.” This provided an opportunity to revisit the framework we shared in October 2023 for navigating the global AI governance landscape.

The pace of institutional and political developments over the past three years has been remarkable. Forums and initiatives that once appeared exploratory have matured into consequential spaces for agenda-setting and norm development, others have shifted in relevance or direction, and yet others have merged or expanded to be more effective and coordinated, as final groups have gone away with little impact. Revisiting our earlier mapping exercise underscored the importance of continuously reassessing where influence is exercised, where implementation is occurring, and where civil society engagement can be most impactful.

To support this ongoing navigation challenge, we are updating our governance grid to reflect how these processes have evolved – and to help stakeholders better identify strategic entry points.

Enabling Accountability of AI Systems

On Tuesday 17th, during “Reinforcements & Learning: A Multistakeholder Convening on AI Governance,” we facilitated a roundtable discussion on “Enabling Accountability of AI Systems.”

A clear takeaway from this exchange was that accountability must operate across multiple, mutually reinforcing layers. Regulatory measures are gaining traction in many jurisdictions, but they are only one piece of the puzzle. Technical standards, impact and risk assessments, transparency and disclosure requirements, auditing mechanisms, and incident monitoring are all emerging as essential tools in translating high-level principles into actionable ways to exercise accountability.

Participants also stressed that accountability cannot be treated as a static compliance exercise. It requires institutional capacity, political will, and coordinated engagement across governance forums. Moving forward demands a more deliberate agenda-setting effort – one that identifies priority issues and enables stakeholders to act coherently across global, regional, and national processes.

Just as importantly, accountability frameworks must confront the structural power imbalances that AI systems can create or exacerbate. Inclusive governance requires not only participation, but responsiveness to the needs of communities disproportionately affected by AI deployment.

From Dialogue to Agenda

Across both days, one message was consistent: the AI governance ecosystem is no longer defined by a lack of principles. Instead, the central challenge lies in operationalisation – translating shared commitments into concrete, enforceable, and rights-respecting outcomes.

At GPD, we take this challenge seriously. We leave these discussions energised to continue working with partners from the Global Majority and beyond to articulate a clear agenda for action – one that can be advanced across multiple governance forums and processes in the months ahead. Meaningful multistakeholder approach is a means to an end: it is a mechanism to ensure that AI governance is inclusive, accountable, and capable of addressing the real-world impacts of these technologies.