Page 19 - The Annual AI Governance Report 2025 Steering the Future of AI

The Annual AI Governance Report 2025: Steering the Future of AI

2.2 Steps towards addressing the AI Divide

AI companies are making efforts to promote global inclusion by collaborating with diverse
stakeholders, expanding research beyond Western-centric contexts, and supporting AI
development in the Global South. These initiatives include partnering with regional universities,
funding localized data collection, and promoting access to open-source tools and educational
resources for underrepresented communities. By doing so, companies aim to reduce bias,
31
enhance cultural relevance, and broaden access to AI technologies worldwide. Despite these
efforts, there are significant limitations. Many inclusion initiatives are still driven by Global North
institutions, often sidelining local voices and reinforcing top-down dynamics. Challenges such
as unequal resource distribution, language barriers, and the dominance of commercial interests
often hinder meaningful engagement.

Dataset Inclusion

Efforts to improve dataset inclusion in AI development focus on expanding the geographic,
linguistic, and cultural diversity of training data to reduce bias and enhance global relevance.
Initiatives include sourcing data from underrepresented regions, incorporating non-English
languages, and capturing context-specific information that reflects local realities. Some
organizations support community-driven data collection and promote open datasets that are
accessible to researchers and developers in the Global South. However, these efforts often
face challenges such as limited infrastructure, uneven data governance, and ethical concerns
around consent and representation. Ensuring meaningful dataset inclusion requires sustained
investment, local collaboration, and safeguards that prioritize fairness and accountability.

Research Labs

AI research labs have begun expanding their global reach by establishing partnerships with
institutions in the Global South, opening satellite offices, and funding regional AI hubs aimed at
fostering local talent and innovation. These efforts are intended to decentralize AI development
and bring more diverse perspectives into research and deployment. Google, Microsoft, and IBM
have established research labs in the Global South, as well as development centers, customer
support hubs, or data centers in these regions. However, the distribution of AI research facilities
remains uneven. In Southeast Asia, lab representation is limited solely to India; in South America,
to Brazil. Sub-Saharan Africa shows slightly more geographic diversity, with AI labs located in
Accra (Ghana), Nairobi (Kenya), and Johannesburg (South Africa). Grassroots AI education and
training initiatives by communities such as Deep Learning Indaba, Data Science Africa, and
Khipu AI in Latin America aim to increase local AI talent. However, inclusion remains limited, as
decision-making power and core research agendas often remain concentrated in the Global
North. Many collaborations still operate within asymmetrical power structures, where local
contributors have little influence over priorities or outcomes.

2.3 Economic Growth and Productivity Gains

Calculations on the economic growth and productivity gains of generative AI rely on two
types of assumptions: task replaceability and new innovation capabilities. Experts have varying

31 Chan, A., Okolo, C. T., Terner, Z., & Wang, A. (2021, February 2). The Limits of Global Inclusion in AI
Development. arXiv.org.

10