Data as the Foundation of AI: Bridging the Governance Gap in the Age of Artificial Intelligence

As we move at an unprecedented pace into the dynamic world of artificial intelligence (AI), a significant rise in the creation of committees, forums, and summits for AI governance is apparent. However, these essential platforms often miss a critical aspect: the role of data governance. This article highlights the crucial, yet frequently overlooked, connection between AI governance and effective data governance.

The current emphasis on AI governance, which addresses various ethical, legal, and societal issues, often neglects the fact that successful AI governance fundamentally depends on the principles and practices of managing data. This lack of attention has led to a disjointed approach, where the data and AI sectors function separately, often unaware of the vital synergy needed between them.

In their article "Interwoven Realms: Data Governance as the Bedrock for AI Governance" Stefaan G. Verhulst and Friederike Schüür explore how closely linked these two areas are. They presented six reasons why AI governance cannot be achieved without a comprehensive data governance framework. By focusing on this intersection, this article aims to emphasize the importance of incorporating data governance more prominently in discussions about AI, leading to a more unified and effective strategy in managing this transformative technology.

Six Reasons Data Governance is Essential for AI Governance:

1.Data Lifecycle Integration with AI: Artificial intelligence is inherently part of the data lifecycle, which data governance oversees. AI governance deals with the lifecycle of AI systems, including their creation, implementation, monitoring, and retirement. Effective AI governance relies on strong data lifecycle governance to ensure AI systems' dependability and relevance.

2.Enabling Responsible AI Development:

• Responsible Data Availability: AI systems largely need data for development. Data governance ensures responsible data availability, determining what data is accessible, to whom, and under which conditions.
• Data Quality and Integrity: The effectiveness of AI models largely hinges on the quality of the training data. Data governance ensures data accuracy and integrity.
• Data Standards and Interoperability: Data governance establishes standards essential for creating large, interoperable datasets necessary for AI.

3.Data Appropriateness and Representativeness: Data governance ensures that AI systems' data is representative of their user base, which is critical for their effectiveness and fairness.

4.Addressing Inherent AI System Issues: Data governance manages compliance with privacy and data protection laws, reducing risks of non-compliance and inherent risks in AI systems.

5.Establishing AI Social License:

• Stakeholder Engagement and Transparency: Data governance emphasizes transparency and diverse stakeholder involvement, crucial for trust in AI.
• Consent and Social License for Data Re-Use: Proper data governance ensures data re-use respects consent and societal expectations, maintaining public trust, especially in sensitive AI applications.
• Dispute Resolution and Redress: Data governance frameworks include mechanisms vital for addressing AI-related disputes and grievances.
• Holistic, Technology-Agnostic Nature: Data governance, being technology-agnostic, offers a holistic framework adaptable to evolving technologies, unlike the more specialized nature of AI governance.

6.Valuable Lessons for AI Governance Implementation: Data governance provides standards and policies that can be applied to AI, offering insights into quality assurance and ethical data use.

Effective AI governance is intrinsically linked to solid data governance. Data governance not only sets the framework for proper data management but aligns these practices with ethical norms, legal requirements, and public expectations. This is increasingly vital as data is leveraged for AI and other advanced analytical technologies.

Despite their obvious interconnection, discussions on AI governance often fail to link it with data governance. This disconnect may slow meaningful AI governance development, hinder responsible AI system creation, and exacerbate inequalities due to asymmetric data availability.

An integrated approach, merging broad data governance principles with specific requirements of technologies like AI and IoT, promises a more balanced and effective governance structure. This approach ensures that each technology's unique aspects are addressed while maintaining a consistent overarching data management and decision-making framework.