Panelists say data, AI and big datasets strain Belmont-era consent and subject definitions

Panelists raised concerns that modern data ecosystems complicate definitions of a research subject and the practice of informed consent. Dr. Ivy Tillman and others noted that genomic results and large datasets can affect family members and communities, while data models trained on diverse datasets may still perform poorly for specific subpopulations.

"If there was an area we could add on to Belmont, it would be around who is a subject," a panelist said, noting that the era of large health datasets and AI requires rethinking consent scope and privacy protections. The group urged research into metrics and evidence-based approaches to measure whether protections actually work, echoing repeated Government Accountability Office prompts for better efficacy evaluations.