Crowell & Moring

Report flags legal risks of AI-generated drug evidence

A Crowell & Moring client alert says probabilistic AI models in drug discovery can produce plausible-but-wrong outputs that affect regulatory submissions, trial integrity and legal liability. It lists five risks: hallucinations, unrepresentative training data, opaque reasoning, input sensitivity and adversarial manipulation of training data, and urges experimental validation, human oversight and preparation for FDA and EU guidance.