AI systems may be fueling ‘digital colonialism’ through indigenous data extraction

Research warns that artificial intelligence may also be reproducing patterns of colonial exploitation in the digital age. Without stronger governance frameworks, AI could become another mechanism through which powerful institutions extract value from marginalized populations while leaving them excluded from the benefits.

A new study titled Preventing AI extractivism: the case for braiding Indigenous data justice with ABS for stronger AI data governance, published in the journal AI & Society, examines this growing risk and proposes a legal and governance framework to prevent it. The research reveals that artificial intelligence systems are increasingly built on data derived from Indigenous communities without meaningful oversight, compensation, or participation.

According to the authors, AI development is increasingly dependent on massive datasets that often include Indigenous linguistic records, ecological knowledge, biometric information, and geospatial data. These resources are frequently collected from online repositories, digital archives, and surveillance technologies under the assumption that publicly accessible data is free to use. The study claims that this approach risks replicating historical colonial extractivism, where valuable resources were taken from Indigenous communities without consent and transformed into economic capital elsewhere.

The researchers frame this phenomenon as AI extractivism, a process in which knowledge and cultural resources embedded within Indigenous societies are converted into digital assets for machine learning systems.

How AI is reproducing patterns of digital colonialism

The study identifies several areas where AI technologies are already replicating historical patterns of colonial resource extraction. One notable example involves the use of Indigenous languages in AI training datasets. Machine learning models designed for speech recognition, translation, and conversational AI increasingly rely on recordings of endangered languages gathered from the internet. These recordings are frequently scraped from online platforms without consultation with the communities that produced them.

For many Indigenous groups, language preservation is deeply tied to cultural identity and historical survival. Languages that were once suppressed through colonial policies are now being digitized and incorporated into AI systems without community involvement. The researchers warn that this process can transform cultural heritage into commercial data resources controlled by external corporations.

Another area of concern is the expansion of biometric technologies powered by AI. Facial recognition systems, fingerprint databases, and large-scale biometric identification programs are becoming central to modern surveillance infrastructure. The study highlights that these technologies often disproportionately target Indigenous communities and activists, particularly in contexts where environmental or land rights conflicts are ongoing.

Biometric surveillance has deep historical roots in colonial governance systems that monitored and categorized colonized populations. The researchers argue that AI-driven biometric tools risk reinforcing similar power dynamics, especially when deployed in large national identification systems or policing strategies that disproportionately affect marginalized groups.

The use of AI in geospatial analysis also raises significant concerns. Advanced machine learning systems now process satellite imagery and remote sensing data to identify mineral deposits, archaeological sites, and environmental patterns. While these technologies are often framed as tools for innovation and resource management, they can also reveal the locations of sacred sites, burial grounds, and culturally significant landscapes.

When such data becomes publicly accessible through mapping platforms or research databases, it can expose Indigenous territories to tourism, commercial exploitation, or extractive industries. The study argues that AI-driven mapping technologies may therefore function as modern extensions of colonial cartography, which historically played a central role in territorial control and resource extraction.

Ecological and biological data mining represents another emerging risk. Advances in digital biology allow scientists and companies to sequence the genetic information of plants, animals, and ecosystems. Once this genetic data is uploaded to global databases, it can be accessed and used anywhere in the world.

The researchers warn that this process may enable new forms of biopiracy. Pharmaceutical or biotechnology companies could use digital genetic data derived from Indigenous ecological knowledge to develop commercial products without compensating the communities that preserved and transmitted that knowledge for generations.

These examples show how AI development can transform cultural knowledge, environmental information, and linguistic heritage into digital commodities. The study argues that this pattern mirrors earlier forms of colonial extraction, where valuable resources were removed from Indigenous territories and integrated into global economic systems.

Why existing AI governance is not enough

Current AI governance frameworks are not equipped to address the ethical challenges posed by data extraction from Indigenous communities. While international agreements exist to regulate the use of biological resources, similar protections are largely absent in the digital domain.

The authors draw attention to the Convention on Biological Diversity and the Nagoya Protocol, two international agreements that regulate access to genetic resources and traditional knowledge. These frameworks introduced the concept of Access and Benefit Sharing (ABS), which requires researchers and companies to obtain consent before using biological materials and to share benefits derived from them.

Under the Nagoya Protocol, users of genetic resources must secure prior informed consent from the communities providing those resources and negotiate mutually agreed terms for their use. The agreements also mandate fair and equitable sharing of benefits, which may include financial compensation, technology transfer, or capacity-building initiatives.

These principles were developed in response to historical controversies involving biopiracy, where corporations patented products derived from Indigenous knowledge without compensation. The study argues that the same logic should apply to digital data used in artificial intelligence systems.

However, the current global AI ecosystem lacks comparable legal requirements. Companies often justify data collection practices by citing the concept of "open data," which treats publicly available information as freely usable for any purpose. According to the researchers, this interpretation ignores power imbalances and fails to recognize that many forms of knowledge embedded in digital datasets originate from communities that never consented to their use.

The authors warn that open data policies, when applied without cultural safeguards, can replicate the colonial concept of terra nullius. Just as colonial authorities once declared Indigenous lands to be unowned and therefore open for exploitation, digital platforms can treat Indigenous knowledge as an unrestricted resource.

This dynamic creates a new version of the commons dilemma, where corporations with advanced technological capabilities extract value from global data resources while the communities that generated the knowledge remain excluded from decision-making and economic benefits.

The study also emphasizes the significance of Indigenous data sovereignty, a growing movement asserting that Indigenous peoples have the right to control data about their communities, territories, and cultures. Indigenous data sovereignty extends the principle of self-determination into the digital realm, emphasizing that data governance should reflect the priorities and values of the communities involved.

Several frameworks have emerged to support this approach. Among them are the CARE Principles, which emphasize collective benefit, authority to control, responsibility, and ethics in data governance. Another influential framework is OCAP, which stands for ownership, control, access, and possession of data.

These principles challenge the dominant model of data governance that prioritizes technological innovation and open access. Instead, they emphasize community authority, ethical responsibility, and equitable distribution of benefits derived from data.

A proposed global framework for ethical AI data governance

To address the risks of AI extractivism, the study proposes a new international governance model that combines Access and Benefit Sharing mechanisms with Indigenous data governance frameworks. The researchers describe this approach as a braided governance model, in which legal, ethical, and operational principles are integrated to create stronger protections for Indigenous data.

In this model, Access and Benefit Sharing provides the legal foundation for regulating how data is accessed and used. The framework would require AI developers to obtain prior informed consent before collecting or using Indigenous data, negotiate mutually agreed terms with communities, and ensure that benefits generated from AI applications are distributed fairly.

The OCAP framework would serve as an operational system defining how data is collected, stored, and shared. Under this approach, Indigenous communities would retain ownership and control over data related to their cultures, territories, and knowledge systems. Access to that data would be governed by community-defined protocols rather than external institutions.

The CARE principles would provide the ethical foundation of the governance model, ensuring that AI development aligns with Indigenous values and priorities. This would require AI projects to demonstrate collective benefit for communities and respect the authority of Indigenous governance systems in decision-making processes.

The study argues that combining these frameworks would create a multi-layered governance structure capable of addressing the complex challenges posed by AI. Legal agreements would establish enforceable obligations, operational frameworks would define practical governance procedures, and ethical principles would guide responsible data use.

The researchers also discuss potential technological tools that could support this governance model. Blockchain systems could be used to create transparent consent registries, ensuring that permissions granted by communities are traceable throughout the lifecycle of AI datasets. Digital watermarking and metadata tagging could embed information about the origin and permitted use of Indigenous data directly into digital files.

Community-controlled data portals could serve as gateways for researchers and companies seeking access to Indigenous datasets. These portals would allow communities to monitor data use, impose restrictions, and ensure that agreements are respected over time.

The study reiterates that AI governance must move beyond voluntary guidelines toward binding international standards. Because AI development operates across borders and cloud infrastructures, domestic regulations alone are unlikely to prevent data extraction from communities in different parts of the world.

An international treaty modeled on the Nagoya Protocol could establish consistent rules for AI data governance, ensuring that companies, governments, and research institutions follow the same standards regardless of where they operate. Such a framework would not necessarily limit technological innovation. Instead, the researchers argue that it could encourage more collaborative forms of AI development in which Indigenous communities become active partners rather than passive data sources.