Closing AI drug-regulatory gaps through harmonized oversight

Proposes a risk-tiered regulatory framework that treats discovery AI and evidence-generating AI differently. Relative to existing frameworks that lack enforceable standards, it stratifies regulatory treatment by the functional role of AI in drug development rather than treating all AI applications uniformly. A position and framework piece based on observation of existing regulatory gaps, without empirical data or case validation.

Requires mandatory impact assessments and Investigational New Drug disclosure when AI influences decisions. Converts expectations that previously remained at the guideline level into specific disclosure and assessment obligations at key regulatory junctures. A normative recommendation grounded in the stated lack of enforceable standards, with no implementation-effect data.

Advocates transforming guidelines into binding, risk-proportionate regulation. Shifts from voluntary guidelines toward enforceable binding oversight, emphasizing that regulatory intensity should match risk level. A policy proposal supported in the text by the background fact of more than 100 AI-assisted pipelines in trials, without quantitative assessment.

The framework addresses drug development and regulatory settings, applying where AI participates in drug discovery or evidence generation and may influence decisions, for reference by regulators, sponsors, and researchers; its design intent is to achieve risk-proportionate oversight through impact assessments and IND disclosure when AI influences decisions.

This is a summary-level text that does not provide operational details, tiering criteria, or implementation cases; readers should still watch how risk tiers are defined, how impact assessments and IND disclosure are executed, and how binding regulation would be implemented across jurisdictions.