Major US and Canadian research institutions are treating AI-generated synthetic patient datasets as non-PHI and waiving traditional ethics board review. The move can shorten clinical research cycles, but it also exposes a governance gap while regulators and institutions still lack consistent rules for synthetic data.
Research centers treat synthetic patient data as “non-PHI,” bypassing ethics review
Several research institutions in the US and Canada are waiving traditional ethics board review for projects that use AI-generated synthetic patient data, arguing that the data does not qualify as personal health information (PHI). The institutions cited include Washington University School of Medicine (WashU Med), Children’s Hospital of Eastern Ontario (CHEO), and the Ottawa Hospital.
The practice reflects a regulatory gray area: synthetic datasets are increasingly used in clinical research and AI prototyping, yet governance expectations and oversight pathways are not consistently defined. The result is a faster path to start studies and iterate on models, alongside new questions about what “ethical review” should look like when the dataset is synthetic but derived from real patient populations.
- Speed is real, but so is accountability. Waiving ethics review can compress timelines for early-stage analysis and AI tool prototyping, but it doesn’t remove the need for internal approval, documentation, and risk sign-off.
- “Not PHI” isn’t the same as “no privacy risk.” Even if synthetic data is treated as non-PHI, teams still need to assess re-identification risk, dataset leakage scenarios, and downstream linkage risk—especially when synthetic records are generated from sensitive clinical sources.
- Expect controls to shift from IRB to engineering. If ethics boards step back, privacy engineers and data governance leads will likely carry more of the burden: access controls, audit trails, reproducibility, and clear rules for who can generate, share, and publish synthetic outputs.
- This is a policy vacuum founders should plan for. Products and partnerships that rely on “synthetic = exempt” assumptions may face friction when reviewers, hospital partners, or regulators tighten interpretations or introduce synthetic-specific guidance.