Governments worldwide play a pivotal role in shaping the AI ecosystem through instruments such as legislation, funding, and international collaboration. However, our recent work reveals significant disparities in how countries emphasise data within their AI policies, potentially worsening the global data divide.
Analysing 512 policy documents from 64 countries, we find that a small group of typically wealthier nations with robust open data practices are more likely to focus on data-centric AI topics. In contrast, low and middle-income countries (LMICs), which generally lack a focus on data-centric AI topics, may find their ability to engage in global AI governance efforts hindered.
Crucial aspects of data governance—such as data lineage, provenance, transparency, licensing, and standards—are frequently overlooked, even by countries that emphasise data overall. Also, importantly, there's a global decline in the focus on open data within AI policies, despite its role in ensuring equitable access and driving the critical mass of data necessary for innovation.
To address these challenges, we recommend:
- Supporting LMICs: International bodies should provide guidance on digital infrastructure and accessing high-quality data resources for LMICs, enabling them to participate fully in global AI governance.
- Investing in data-centric tools: Policymakers should adopt robust data documentation practices and data-centric AI toolkits, supported by machine-readable metadata, to enhance transparency and regulatory oversight.
- Promoting equitable data sharing: There's an urgent need to encourage safe and equitable data sharing, reconsider open data policies, and explore data access remedies specific to AI contexts. These challenges are further explored in our forthcoming research "The UK Government as a Data Provider for AI".
This work aims to provide valuable insights and resources to guide responsible AI development and ensure that the benefits of AI are accessible globally. We invite policymakers, researchers, and all stakeholders to engage with our findings and collaborate on shaping a more equitable AI ecosystem. Through this collaboration, can support the development of AI systems that are innovative, transparent, and beneficial to all societies by prioritising data-centric approaches in AI policy.
