CARE as a Model of Technological Refusal: Collective Data Sovereignty in the Age of AI
Introduction
In recent years, artificial intelligence (AI) governance is often primarily framed through the lens of one question: how can this new technology be used responsibly. Governments, corporations, and international organizations focus on problems such as algorithmic bias, transparency, accountability, and risk mitigation. These areas of focus are useful and vital for the communities that choose to opt into its implementation and execution. However, just as valid is the right of refusal and the protections built into such a choice. From this concern, an important question emerges: who has the authority to refuse data extraction and AI use altogether?
The CARE Principles for Indigenous Data Governance offer a different answer to that question. Developed by Indigenous scholars and institutions through the Global Indigenous Data Alliance, CARE reframes data and digital governance as a matter of communal autonomy rather than technical optimization. CARE stands for Collective Benefit, Authority to Control, Responsibility, and Ethics. Unlike Western privacy models that emphasize individual rights and procedural safeguards, CARE centers joint governance over whether data may be collected, circulated, or used at all.
In this paper, CARE is presented as a case study in refusal-based technological governance. Rather than treating CARE as merely a noble supplement to existing AI practices, this paper suggests that CARE delineates a distinct political model of governance rooted in collective refusal, mutual responsibility, and sovereign authority. Through an analysis of CARE’s origins, principles, and current applications in AI-related contexts, this paper explores how CARE challenges the dominant assumption that marginalized communities must participate in digital systems in order to be protected. At the same time, the paper critically examines the limits of CARE as it confronts the systemic power of national governments, global markets, and corporate technology infrastructures.
By analyzing CARE as a viable governance option rather than an abstract theory, this paper illustrates refusal not as technological de-progession but as a standard and viable form of political authority in the age of AI.
Origins of CARE and the Political Context of Indigenous Data Governance
Going back centuries, Indigenous communities across the globe have been treated as objects of study rather than as political authorities over their own cultural knowledge systems. On par with the historical practices of colonialism, governments, universities, medical researchers, and private corporations collected biological samples, cultural artifacts, geographic data, oral histories, and demographic information without meaningful consent or reciprocal benefit.
In the late twentieth and early twenty-first centuries, this method of extraction intensified under what critics describe as “data colonialism ³.” Large-scale digitization projects, open data initiatives, and emerging AI systems drastically expanded the technological wherewithal to collect, duplicate, and circulate information about Indigenous peoples and their territories. Satellite imaging, biometric identification systems, genomic databases, and machine learning models increasingly squandered Indigenous life into globally mobile data assets. Crucially, this expansion often occurred under the guise of openness, innovation, and universal benefit—while Indigenous communities remained systematically excluded from control over how their data were used or monetized.
In response, Indigenous scholars and governance bodies began conjuring formal data autonomy frameworks. These efforts dictated that data about Indigenous peoples, lands, and cultures are not neutral resources. Indigenous data sovereignty asserts that Indigenous nations have inherent rights to govern the creation, ownership, access, and application of data derived from their communities.
The CARE Principles were formally introduced in 2019–2020 by members of the Global Indigenous Data Alliance as a counter-framework to dominant open data and “ethical AI” models ¹. CARE was designed to complement, but also challenge, the FAIR data principles (Findable, Accessible, Interoperable, Reusable). These principles prioritize technical efficiency and data circulation. Where FAIR centers the movement of data through systems, CARE caters to the rights, responsibilities, and collective authority of the people to whom that data pertains.
From its beginnings, CARE has been expressed in explicitly political terms. It intervenes at the level of governmental authority, advocating that Indigenous communities should maintain decision-making power over whether data is collected and how it is used. This shift from procedural ethics to sovereign authority is what makes CARE particularly significant for AI governance.
The Four CARE Principles as Operational Governance
The CARE framework is built around four connected ideas: Collective Benefit, Authority to Control, Responsibility, and Ethics. These ideas explain how CARE aims to treat data governance as part of everyday political life.
Collective Benefit
Collective Benefit asks: who is actually helped by the use of data? Throughout history, data taken from marginalized communities has mainly served people and institutions outside of those communities such as researchers, corporations, and government agencies. CARE seeks to undermine this model. Under this principle, data use is only justified if it clearly supports the well-being of the community it comes from. That can look like better healthcare, protection of land and water, cultural preservation, or stronger self-determination.
The purpose here is that benefit is defined at the collective level. Under this idea, individual consent is not treated as enough. A person might technically agree to share their data, but that same data could still be used in ways that harm the larger community i.e. enabling surveillance, reinforcing harmful narratives, or supporting new forms of extraction. CARE treats benefit as something that the community must be able to recognize and claim, not just something an individual signs off on which is an important distinction.
Authority to Control
Authority to Control is the idea that Indigenous communities never gave up their right to govern data about their people, lands, and identities. That authority includes the power to decide whether data is gathered at all, the ability to withdraw it later, and the right to limit who can use it and for what purpose.
This approach is contrary to Western legal frameworks that treat data as a commodity or as an abstract informational resource. Under CARE, governance does not get handed over to outside institutions through technical oversight alone but, instead, remains grounded in Indigenous political autonomy. In the context of AI, this means communities retain the right to block their data from being used in training datasets, surveillance systems, or predictive models—even when those systems are presented as innovative or beneficial to the public.
Responsibility
The principle of Responsibility is about what happens after data is collected. CARE understands data governance as a relationship, not a transaction. The people and institutions that manage data are responsible not only for keeping it secure, but also for staying accountable to the communities it comes from. That means ongoing communication, transparency, and a willingness to respond when concerns are raised.
Responsibility also stretches across time. Decisions made about data today can shape community life decades into the future. CARE places weight on that long horizon. It asks that governance choices be made with future generations in mind, rather than being driven only by short-term research goals or market incentives.
Ethics
Ethics, within CARE, is not treated as a universal checklist developed in corporate or academic settings far from the communities most affected by data extraction. Instead, ethical governance has to align with the values, laws, and worldviews of the people whose data is being used. This shifts ethics away from being something applied from the outside and toward something that grows out of lived community standards and responsibilities.
Taken together, these four principles form a governance system that centers collective authority, shared responsibility, and the possibility of refusal. CARE does not privilege technical efficiency or seamless data circulation. It prioritizes relationships, limits, and political decision-making. In the context of AI, this positions CARE as a direct challenge to governance models that assume participation is inevitable and that optimization is the ultimate goal.
CARE Versus Western AI Governance Frameworks
Most contemporary AI governance frameworks operate within a managerial or regulatory paradigm ². Examples include data protection laws such as the General Data Protection Regulation (GDPR), corporate AI ethics guidelines, and government AI risk-management strategies. While these frameworks differ in scope and enforcement power, they share several foundational assumptions: that AI systems will be built and deployed; that data will continue to circulate through global markets; and that governance consists primarily in mitigating risks associated with those deployments.
Western privacy regimes are typically grounded in individual rights. Consent is framed as a contractual agreement between a data subject and a data controller. However, in practice, consent is often coerced through disproportionate power relations, nontransparent terms of service, and the perceived social necessity of global digital participation. Moreover, individual consent frameworks are poorly suited to addressing collective harms such as cultural misrepresentation, group profiling, or community-wide surveillance.
Corporate AI ethics initiatives similarly focus on principles such as fairness, transparency, explainability, and accountability ⁴. While these principles address important technical failures, they rarely question the broader political economy of AI. Ethics becomes a procedural layer applied after the fact, rather than a challenge to whether certain systems should exist at all.
CARE diverges from these models with an important distinction. It does not frame governance as a matter of technical compliance or procedural consent. Instead, it frames governance as a question of political authority: who decides whether data may be extracted and used in the first place. Where Western frameworks manage the risks of AI use, CARE emphasizes the importance of retaining the option to refuse participation in AI systems altogether.
This distinction is especially significant in the context of marginalized communities. Under dominant AI governance regimes, marginalized groups are often positioned as passive subjects of protection rather than as active political authorities. CARE reverses this relation by situating communities as sovereign decision-makers rather than as objects of regulation.
CARE in the Age of AI
Although CARE originated in the context of research governance and data stewardship, its relevance has expanded rapidly in response to AI-driven data extraction. AI systems depend on large-scale, aggregated datasets that often include sensitive cultural, biological, environmental, and behavioral information. Machine learning models trained on such data can reproduce group-level harms even when individual privacy is technically preserved.
CARE has increasingly been applied to areas such as:
Healthcare AI – Indigenous health data has historically been exploited for genomic and epidemiological research without returning meaningful benefits to communities. CARE-based governance requires that Indigenous nations retain authority over whether their health data are used in predictive models, diagnostic tools, or pharmaceutical research.
Environmental and Climate Modeling – Satellite imagery, geospatial datasets, and ecological monitoring tools increasingly inform climate policy and resource extraction. CARE asserts that data about Indigenous lands cannot be freely incorporated into models that facilitate external exploitation without community authorization.
Digital Identity and Biometrics – Facial recognition, biometric ID systems, and population registries pose particular risks to Indigenous communities, given histories of surveillance, displacement, and forced assimilation. CARE provides a governance basis for refusing biometric data collection altogether.
In each of these domains, CARE can operate as an anti-extractive constraint on certain forms of AI deployment. Rather than asking how to make AI systems fairer within existing political economies, CARE asks whether such systems should be permitted at all when they undermine digital sovereignty.
Authority Without Technical Literacy
One of the most significant implications of CARE is that it separates technological governance from technical expertise. In current prevailing policy discourses, effective participation in AI governance is often assumed to require advanced technical knowledge. This assumption advantages already powerful institutions—corporations, research universities, and state agencies—while rendering marginalized communities “deficient” participants in governance.
CARE rejects this premise. It establishes that governance authority does not derive from technical skill but from political and relational autonomy. Communities do not need to understand the inner workings of machine learning algorithms in order to assert boundaries over data extraction. What matters is not how a model functions, but whether a community has authorized its creation and use.
This distinction has important implications for fair and shared governance. By affirming that authority precedes participation, CARE empower communities to exercise power even in the absence of technical capacity. Technical knowledge may still be valuable, but it is no longer the gatekeeper of sovereignty.
The Limits and Constraints of CARE
Despite its radical implications, CARE still operates within real-world political constraints. Several limitations shape its current impact.
First, CARE depends heavily on institutional recognition. Its effectiveness is strongest in contexts where Indigenous governance structures are legally acknowledged by states or international bodies. Where such recognition is weak or contested, CARE-based refusals may be overridden by national data policies or corporate interests.
Second, CARE currently functions most clearly in research and institutional data settings. Its application to commercial AI systems remains uneven and contested. Global technology corporations are not legally bound by CARE unless governments incorporate its principles into binding law.
Third, CARE faces challenges of enforcement at scale. Data flows across borders at speeds and volumes that exceed the regulatory capacity of many communities and states. Even when communities refuse consent, technical enforcement mechanisms may be difficult to sustain against powerful data infrastructures.
These limits do not invalidate CARE, but they reveal the dissonance between community governance and global digital capitalism. CARE represents a political claim to authority that still must contend with entrenched systems of power.
Sources and Method
This case study draws on a combination of primary and secondary sources. Primary sources include the original formulation of the CARE Principles by the Global Indigenous Data Alliance and Carroll et al., along with institutional CARE documentation. Secondary sources include scholarship on data colonialism, extractive digital infrastructures, and Western AI governance frameworks, including works by Kate Crawford, Nick Couldry and Ulises A. Mejias, and Cathy O’Neill. Together, these sources support a comparative analysis of refusal-based governance and dominant regulatory models.
Conclusion
CARE represents a necessary departure from dominant models of AI governance. Where most frameworks seek to manage the risks of AI deployment, CARE asserts a more fundamental political right: the technological autonomy to refuse data extraction and algorithmic implementation altogether. By centering sovereignty, relational accountability, and collective benefit, CARE reframes data governance as a question of power rather than optimization.
As a case study, CARE demonstrates that technological governance need not be grounded in participation, innovation, or technical proficiency. Instead, it can be grounded in refusal, boundary-making, and sovereign authority. At the same time, CARE’s limits reveal the formidable structural power of states and corporations in the global AI economy.
Consequently, CARE shifts the central question of AI governance from “How should we use AI?” to “Who has the authority to decide whether AI may be used at all?” In doing so, it exposes the deeply political nature of tech systems and offers a framework through which refusal may be understood as a legitimate and organized form of governance.
The author acknowledges the use of OpenAI's ChatGPT (Version GPT-4o) to improve the readability and structural flow of the Introduction and Discussion sections. The tool was utilized strictly for paragraph tightening and grammatical refinement. All conceptual frameworks, data interpretations, and final editorial decisions were made and verified entirely by the human author, who accepts full responsibility for the content.
Notes
Stephanie Russo Carroll et al., “The CARE Principles for Indigenous Data Governance,” Data Science Journal 19, no. 1 (2020): 1–12.
Kate Crawford, Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence (New Haven, CT: Yale University Press, 2021), 1–22.
Nick Couldry and Ulises A. Mejias, The Costs of Connection: How Data Is Colonizing Human Life and Appropriating It for Capitalism (Stanford, CA: Stanford University Press, 2019), 33–55.
Cathy O’Neill, Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy (New York: Crown, 2016), 83–120.
Global Indigenous Data Alliance, “CARE Principles for Indigenous Data Governance,” 2020, https://www.gida-global.org/care.
Bibliography
Carroll, Stephanie Russo, et al. “The CARE Principles for Indigenous Data Governance.” Data Science Journal 19, no. 1 (2020): 1–12.
Crawford, Kate. Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence. New Haven, CT: Yale University Press, 2021.
Couldry, Nick, and Ulises A. Mejias. The Costs of Connection: How Data Is Colonizing Human Life and Appropriating It for Capitalism. Stanford, CA: Stanford University Press, 2019.
O’Neill, Cathy. Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy. New York: Crown, 2016.
Global Indigenous Data Alliance. “CARE Principles for Indigenous Data Governance.” 2020. https://www.gida-global.org/care.
