Work

An Ethical Framework for Safety and Governance in Biomedical Self-Driving Laboratories: Challenges and Perspectives

Year: 2026
Type: article
Abstract: Self-driving laboratories close the loop between hypothesis generation, experimental execution, and analysis, removing the human from the inner cycle of discovery. Their migration from chemistry and materials science into biomedicine introduces ethical problems that the chemistry literature does not encounter: experiments run on human-derived material, generate data about identifiable people, produce candidates intended for patients, and act on living systems whose variability can masquerade as signal. We argue that biomedical self-driving laboratories defeat the categorical triggers on which biomedical oversight depends, because what an autonomous platform will do is determined at runtime by a model rather than at review time by an investigator. We propose Proportionate Autonomy Governance, a framework that replaces protocol review with envelope review: oversight bodies approve a formally specified and machine-checkable region of experiment space together with the invariants that constrain search within it. Five duties, bounding, tracing, controlling, verifying, and answering, scale jointly with a platform’s autonomy level and consequence tier, and we specify what each requires at each position in that grid. We work the framework through a hypothetical induced pluripotent stem cell differentiation campaign, giving its envelope, autonomy annex, and accountability register, and we state what the framework costs and what it does not fix. We map current instruments onto this framework and identify where they fail to bind, and we relate the proposal to operational design domains, runtime assurance, and safety cases in established safety-critical practice. We report a search for documented safety incidents in autonomous laboratories, which returned none, and we are explicit about the limits of what that absence supports. We close with an implementation agenda.
Language: English
Field: Nursing
Open Access status: hybrid
OpenAlex ID: W7236154558
FWCI:
Cites: 113
Cited by:
Related to: 10