Laying the foundations for a collective approach to health data

L’Initiative IA + Société est ravie de continuer à amplifier les voix des chercheurs émergents du monde entier grâce à notre série de blogues bilingues « Voix émergentes mondiales en IA et société » qui font suite aux ateliers mondiaux des jeunes chercheurs en IA et régulation et de la conférence Shaping AI for Just Futures. Cette série met en avant les recherches innovantes présentées lors de ces événements – et au-delà – afin de faciliter le partage des connaissances et d’engager un réseau mondial dans un dialogue interdisciplinaire constructif.

Le blog suivant a été rédigé par Matheus Falcão qui a été sélectionné par un comité international pour présenter son affiche  « Governance of Health Data for AI Innovation: Reconciling personal data protection, economic justice and non-discrimination in the context of Brazil », préparée avec Renan Gadoni Canaan, dans le cadre de la session d’affiche de la conférence Shaping AI for Just Futures et dont le blog s’inspire.

Cette contribution scientifique n’est disponible qu’en anglais.

Why a collective approach on health data?

Health data provides an immense value to AI innovation, as it is one of the key components for training machine learning-based AI models. For this piece, I consider health data as at least three different types: personal health data, which carries information about the health state of an identifiable person; epidemiological data, which contains information about the health state of a population; and data about underlying conditions of health, such as data about pathogens.

The more data one can gather means greater power for developing AI tools applicable to health, which put large competitors, such as big technological or pharmaceutical companies, and even countries, in a position of advantage, making it harder for competitors to catch up. 

This leads to market concentration, often translated into an immense amount of political power and control over health systems and health innovation. Health data, however, carries a collective nature, since it is an extension of human identity. 

The challenge of building a collective approach for health data

In the context of the digital economy and the advancement of AI tools, there is a strong push for the free flow of data, through interoperability, open-access models and other mechanisms for data collection and sharing. This is indeed an important goal, for instance, to avoid the development of siloed AI models, which are trained with datasets restricted to a specific location or institution (e.g., a single hospital). 

The main legal frameworks, however, are so far unable to resolve the issue of acknowledging the value of those datasets through properly sharing the economic benefits that come out of AI innovation.

The main legal frameworks for data or AI regulation are either individualistic approaches to protection, such as those inspired by the General Data Protection Regulation (GDPR), or based on risk or impact assessments of AI systems such as the recent EU’s AI Act or on the regulation of AI-enabled medical devices. Neither approach considers the economic aggregated value of health data, which could be addressed by a mechanism to share the benefits of data-driven innovation.

Benefit sharing in the context of Global Health: potentials and limitations

The unfinished negotiations over the Pandemic Treaty in 2024 well illustrate the matter. The treaty was not concluded due to major disagreements and opposing perspectives between Global North and Global South countries. One of those main clashes was around the topic of a Pathogen Access and Benefits Sharing system (PABS). 

Countries from the Global North, in a position synthesized by the European Union, promoted the ample sharing of data, which also unleashed a big potential for AI innovation in drug discovery and early epidemic detection. Under the premise of the emerging “One Health” concept, the position advocated for countries to structure health surveillance initiatives for collecting and sharing epidemiological, environmental and pathogen data.

On the other hand, countries from Global South pledged to include the PABS system, which would basically make the technologies developed out of this data, including vaccines, to be shared with the world, acknowledging the collective value of the data that generated it. In other words, it would make it closer to a public good.

The concept of benefit sharing emerged from the Global Health landscape first with the Nagoya Protocol (2012), a supplement of the 1992 Convention of Biological Diversity (CBD), which lists a few examples of how technology can be shared, including prices and sharing of intellectual property (IP) rights with traditional communities. However, the non-adherence of the United States to the CBD and the opposition from big companies rendered this tool considerably toothless. The Pandemic Influenza Preparedness (PIP) Framework, however, is a concrete example that put this idea into practice.

The PIP Framework is a benefit-sharing mechanism for technologies generated with samples of the influenza virus that were shared with the World Health Organization (WHO). It is a structure hosted by the WHO that makes it mandatory for companies that want to use virus samples (or respective data) to pay an amount. As interesting as this innovative structure is, its creation emerged from a political dilemma generated by the refusal of the Indonesian government to share samples with the WHO in 2006, under the justification that this effort was not rewarded and Low- and Middle-Income Countries would not be able to access the vaccines from it due to high prices and IP rights protection.

In spite of its success, the PIP Framework is still highly restricted to a single disease. In the era of AI, where the potential of producing value from different types of health data and applying it on tools such as early detection systems and drug development, there is much more room for discussion in that regard. 

Towards a new collective approach for data and AI

The PIP Framework and the Nagoya Protocol are examples of ways to shed light into a wider and relevant discussion in the AI era: how to create a legal framework that acknowledges the collective approach of health data and make innovation equitable and accessible to the communities that generate the data by simply existing. 

The Brazilian Association of Collective Health (Abrasco), one of the first in the field of health informatics and information in the country, states that health information is a patrimony of the whole society that registers its clinical, epidemiological, genetic and cybercultural memory. In the age of AI, where health data can be an engine for economic development, this perspective can also use law and regulation to help distribute this development with the whole population.

Future discussions on this topic could explore the creation of public infrastructures dedicated to storing and processing data. This could include government-owned data centers for epidemiological data, equipped with robust governance mechanisms. Additionally, fostering international collaborations—aligned with principles like those underpinning the PIP Framework or the proposed Pandemic Treaty incorporating the PABS system—might prove crucial.

However, there is likely a more profound challenge for the field of law: the need to develop innovative frameworks that shift from proprietary health data regimes to collective ones. This transition could redefine how health data is governed and shared globally. Undoubtedly, this is a complex issue worthy of deep consideration, but it represents yet another critical item on the broader agenda of AI regulation.

Abrasco (2013), 2º Plano Diretor para o Desenvolvimento da Informação e Tecnologia de Informação em Saúde ABRASCO. 

Fidler, David P., and Lawrence O. Gostin (2011). The WHO pandemic influenza preparedness framework: a milestone in global governance for health. 306.2 Jama 200-201.

Gurumurthy, Anita, and Nandini Chami (2022). Governing the resource of data: to what end and for whom? Conceptual building blocks of a semi-commons approach. Conceptual Building Blocks of a Semi-Commons Approach.

Viljoen, Salomé. (2021) A relational theory of data governance. 131:2 Yale LJ  573. 

World Health Organization (2024). Ethics and governance of artificial intelligence for health: large multi-modal models. WHO Guidance.

Matheus Z. Falcão is an Alex Trebek Visiting Doctoral Student Fellow on AI and Society with the AI + Society Initiative at the Centre for Law, Technology and Society at the University of Ottawa. PhD Candidate at the University of São Paulo, Brazil. His current work focuses on Digital Health, Global Health, and Health Systems. He is the director of the Brazilian Centre for Health Studies (Cebes) and an associate researcher at the Health Law Research Centre of the University of São Paulo. He has experience with research projects in partnership with the Pan American Health Organisation (PAHO/WHO), the National Council of Justice of Brazil and other public offices and civil society organizations, including the Brazilian Institute for Consumers Defense (Idec) and the South Centre.

This content is provided by the AI + Society Initiative to help amplify the conversation and research around the ethical, legal and societal implications of artificial intelligence. Opinions and errors are those of the author(s), and not of the AI + Society Initiative, the Centre for Law, Technology and Society, or the University of Ottawa.