Industrial Policy and Data Extractive Industries

Industrial policy is back. The Point Lisas Policy best exemplifies the range of choices associated with industrial development during the first Golden Age of expansion, from 1974 to 1982. However, global industrial policy today has made the extractive-led development agenda of Artificial Intelligence (AI) its centrepiece.

Industrial policy is now enfolded in new techniques, tools, and the constellations of beliefs and practices concerning extractive-led development in data industries. Proponents of the new attitudes discuss industrial policy in broader terms rather than the old import-substitution strategies or the inward-looking ideas of the new national populism.

As the vanguard of European colonisation, extractive industries provided capital and resources to support “wealth extraction enclaves.” Today, industrial policy convenes a broader church of ideas and people. The new façade of industrial policy turns towards investments in alternative governance models which uproot extractive data practices and elevate the rights and interests of low- and middle-income country (LMIC) communities.

These new industrial policy paradigms enable AI to be yoked as a tool for collective advantage, rather than one that deepens global inequalities, and widens regional and domestic intergenerational immobility and inherited inequality. The transnational nature of AI development clouds the enforcement of protections. Many AI models rely on aggregated, cross-border data, making it difficult to ensure compliance with licensing terms or safeguarding community interests.

Emerging licensing frameworks, like the Montreal Data License, the Responsible AI License (RAIL), the Allen Institute for AI (AI2) ImpACT License, and the Nwulite Obodo Open Data License, try to incorporate ethical restrictions and innovative governance and accountability mechanisms. However, these approaches are difficult to scale, especially when data providers have limited leverage over AI developers.

Indigenous Data Sovereignty (IDS) presents another model that recognises data as a cultural and collective resource rather than an individual asset. IDS frameworks emphasise self-determination, cultural preservation, and reciprocity, enabling indigenous communities like the Santa Rosa First Peoples Community to retain control over the collection, use, and application of their data and knowledge of plants.

To address data governance challenges, a partnership of The GovLab and the Agence Française de Développement (AFD) has produced a more equitable, participatory approach to data governance in AI ecosystems. The outcome is a report titled “Reimagining Data Governance for AI: Operationalising Social Licensing for Data Reuse.”

The report introduces social licensing as a practical, community-centred model for governing data reuse. Digital technologies are increasingly important in the Orange Economy as they enable new ways to create, distribute, and consume creative content like music, kinetic sculpture, and costume design of the Trinidad and Tobago Carnival.

Designers everywhere are now using AI to generate ideas, explore design options, and even create finished designs that reflect the styles, silhouettes, and palettes of designers elsewhere. This is done using AI image generators like DALL-E, Midjourney, and Stable Diffusion. These AI image generators can transform text descriptions or image uploads into visual representations.

Traditional data governance tools are based on static one-time consent models. However, social licensing supports an ongoing process of engagement, negotiation, and accountability. It repositions data governance as a shared responsibility. This empowers communities and small island developing states to define acceptable uses of their data. This is a step toward ensuring that preferences are honoured throughout the AI lifecycle.

As AI assemblages become more reliant on data from low- and middle-income countries (LMICs), elementary questions surface about who controls these data streams and who benefits from their use and reuse. In many cases, the people and communities who generate this data lose control of its use over time as the data moves with a momentum of its own, with few mechanisms for recourse when harms occur.

Traditional governance tools, especially consent-based frameworks, have grave limitations in the context of AI and large-scale data reuse. Consent models aim to give individuals control over their data and are rooted in notions of provenance. In the context of creative content, “provenance” refers to the documented history of ownership, custody, or location of a work of art, from its creation to its current location. It essentially tracks the “life” of an artwork.

While efforts to improve data provenance, including metadata documentation standards, content authenticity techniques, and opt-in/opt-out registries, offer incremental improvements, they still share many of the same limitations as consent frameworks. These tools fail to account for collective governance and remain difficult to enforce at scale. Without mechanisms for community oversight, compliance, and accountability, provenance tracking alone is insufficient for ensuring responsible AI data governance.

Traditional data governance tools do not account for the collective nature of many datasets or the ways data flows across global AI ecosystems long after initial collection. Consent mechanisms tend to be transactional, static, and focused on individual decision-making, failing to reflect the ongoing, collective governance needs of communities. Consent frameworks also require individuals to understand highly complex data ecosystems and predict future uses of their data.

These models are not designed for AI contexts, where data can be recombined, repurposed, and reinterpreted in unpredictable ways. A social license for data reuse is meant to embed community preferences into enforceable agreements that help ensure data governance reflects social, ethical, and cultural considerations alongside legal requirements.