GA4GH launches new Work Stream to support responsible AI in genomics and health

7 Apr 2026

The Global Alliance for Genomics and Health (GA4GH) is introducing a new Work Stream focused on developing governance recommendations and data standards to foster responsible integration of artificial intelligence (AI) in genomics and health.

By Jaclyn Estrin, GA4GH Senior Science Writer

Artificial Intelligence (AI) stands to increase the productivity and accessibility of scientific discovery in genomics and health. However, siloed approaches to AI risk creating systems that are fragmented, inaccessible to lower-resourced institutions, and unable to interoperate. To address these risks in the context of human health and medicine, the Global Alliance for Genomics and Health (GA4GH) is launching a new AI Work Stream to bring together world experts and coordinate a standardised approach to integrate AI into genomics in a responsible, interoperable, and equitable manner.

GA4GH Work Streams drive the development of genomics and health data standards, policy frameworks, and implementation resources. For the first time since the Work Stream structure launched in 2017, GA4GH is now introducing a new, foundational Work Stream, joining Regulatory & Ethics (REWS) and Data Security (DSWS) in addressing cross-cutting needs within the global genomics community. The new AI Work Stream will be led by Marc Fiume (Co-Founder and CEO, DNAstack) and Susheel Varma (Chief Data Officer, Sage Bionetworks) to provide technical leadership and guide standards development within this space.  

Fiume has contributed to GA4GH for over a decade, serving as a founding Co-Lead of GA4GH’s Discovery Work Stream and launching the Beacon API to develop standards that facilitate genomic data discoverability and sharing. His professional background includes implementing GA4GH standards as part of production systems that power international collaborative research in rare disease, neuroscience, oncology, and infectious disease. Fiume has experience building GA4GH compliant systems that support applications in AI, including model training, federated learning, and agentic science.

Varma has been a longstanding contributor to the Cloud and Discovery Work Streams, focusing on secure data access and analysis. Before joining Sage Bionetworks, he served as the Head of AI and Data Science at the UK Information Commissioner’s Office and was previously Chief Technology Officer at Health Data Research UK, roles centred on building national-scale research infrastructure for ethical sharing of personal health data, as well as regulating how AI systems use sensitive data.

“We have a generational opportunity to understand the complex causes of disease with molecular precision. GA4GH standards are making it possible to make data interoperable across institutions, indications, and borders,” said Fiume. “And with AI, we stand to make scientific discovery exponentially more productive. This Work Stream will help ensure we do this safely, securely, and equitably.”

AI is fundamentally changing how science is being conducted at every stage of the research cycle, from hypothesis generation to analysis and interpretation of results. There is growing interest in the use of AI across the genomics and health ecosystem to understand how the technology can be used to drive efficiencies in scientific research, speed up diagnostic timelines, and advance medical care.

“Genomics only transforms health if the world agrees on how to share, interpret, and govern data,” said Sasha Siegel, Chief Product Officer of GA4GH. “We have been building technical standards and policy frameworks that allow genomic data to move responsibly across institutions and borders. If AI is going to transform genomics in healthcare, it has to scale with integrity, and our role at GA4GH is to make sure that it does.”

The AI Work Stream will develop guidance and health data standards that interoperate across the healthcare ecosystem. This work will be grounded in four key principles:

1. Transparency around how systems are built and on what they depend
2. Replicability across datasets, labs, and populations
3. Governance and ethics that treat equity as a core scientific requirement
4. Effectiveness measured by evaluation, evidence, and real-world impact  

Varma said, “We are trying to galvanise the community to think through what end-to-end AI-driven science projects look like when built on standards: identifying what is missing, and what complementary capabilities we need to add to augment that standard to be AI-native or AI-compatible.”

The Work Stream is designed to be a collaborative forum, welcoming contributors with diverse expertise, ranging from technical and policy to implementation experience. The leadership team aims to engage in cross-Work-Stream collaboration to develop jointly owned products that address shared problem statements.

Fiume said, “GA4GH brings together world experts who think deeply about how to translate the value of human data into better health outcomes. It is the right community to lead the conversation on responsible AI, and to make sure the standards we develop are practical, useful, and widely adopted.”

As their work commences, the Work Stream will produce deliverables including standards, metadata schemas, provenance requirements, conformance tests, implementation guides, reference implementations, position papers, and white papers. The leads are looking to develop implementation demonstration projects that showcase how AI can be used to advance a specific use case with increased efficiency and productivity, while maintaining security and equity.

AI is a transformative technology that has an immense potential to shape the future of genomics and health, increase productivity, and drive discoveries. Yet, it is crucial to harness and steer the use of AI through a collective, international approach to data governance and standards. Varma said, “The trust infrastructure that governs how data is shared, accessed, and used is what determines whether AI systems are safe and equitable in practice, not just in principle. That is what GA4GH does better than anyone, and it is what this Work Stream will build on to foster safe, responsible, and equitable science at scale.”