Poster ID
P-20
Poster Title
GA4GH-RegBot: An AI-Assisted Regulatory Navigation Assistant Complementing the GA4GH Regulatory and Ethics Toolkit for Cross-Border Genomic Data Sharing
Authors
Dedy E. Lingga (National University of Singapore / GA4GH REWS); Jimmy Payyappilly (EMBL-EBI / GA4GH); Ruiyang “Ringo” Chen (New York University)
Abstract
Responsible cross-border sharing of genomic and health-related data requires navigating fragmented, jurisdiction-specific regulatory frameworks — GA4GH guidance, REWS materials, GDPR, and regional data-protection and biobanking law. Locating the applicable provisions, institutional contacts, and accountability pathways for a given study is slow and error-prone, and missed provisions carry legal and ethical risk — a barrier felt acutely in lower-resourced and under-represented settings.
GA4GH-RegBot is an open-source, AI-assisted regulatory navigation assistant developed as a Google Summer of Code 2026 project. Given researcher-supplied study materials, it retrieves applicable provisions from a curated regulatory corpus using hybrid retrieval (dense embeddings and BM25, merged by reciprocal rank fusion) with jurisdiction-aware filtering, and returns a structured, citation-grounded navigation document linking relevant clauses, sources, and suggested governance contacts.
Reliability is enforced in code: recommendations may cite only retrieved passages, are checked for genuine textual support, and low-confidence cases are flagged for human review rather than answered speculatively. Critically, RegBot surfaces evidence for human governance review by Data Protection Officers, Institutional Review Boards, and Data Access Committees; it does not issue compliance determinations or provides legal opinions, and decision authority remains with accountable reviewers. RegBot complements the GA4GH Regulatory and Ethics Toolkit rather than replacing human committees, and processes only regulatory texts and researcher-provided documents — never patient-level data.
A running end-to-end prototype spans multiple GA4GH, and Asia-Pacific jurisdictions, set underway. RegBot demonstrates how retrieval-grounded AI can accelerate responsible cross-border genomic research while preserving human governance authority.
GA4GH-RegBot is an open-source, AI-assisted regulatory navigation assistant developed as a Google Summer of Code 2026 project. Given researcher-supplied study materials, it retrieves applicable provisions from a curated regulatory corpus using hybrid retrieval (dense embeddings and BM25, merged by reciprocal rank fusion) with jurisdiction-aware filtering, and returns a structured, citation-grounded navigation document linking relevant clauses, sources, and suggested governance contacts.
Reliability is enforced in code: recommendations may cite only retrieved passages, are checked for genuine textual support, and low-confidence cases are flagged for human review rather than answered speculatively. Critically, RegBot surfaces evidence for human governance review by Data Protection Officers, Institutional Review Boards, and Data Access Committees; it does not issue compliance determinations or provides legal opinions, and decision authority remains with accountable reviewers. RegBot complements the GA4GH Regulatory and Ethics Toolkit rather than replacing human committees, and processes only regulatory texts and researcher-provided documents — never patient-level data.
A running end-to-end prototype spans multiple GA4GH, and Asia-Pacific jurisdictions, set underway. RegBot demonstrates how retrieval-grounded AI can accelerate responsible cross-border genomic research while preserving human governance authority.
Digital Poster