A small-molecule therapy whose structure was proposed end-to-end by a generative model has cleared Phase 3 trials — a milestone the industry has been chasing for a decade.
The US Food and Drug Administration has granted full approval to Insilico Medicine's INS018_055, a treatment for idiopathic pulmonary fibrosis whose entire molecular structure was generated by an AI model. It is the first time a fully AI-originated small molecule has cleared the agency's rigorous Phase 3 bar.
The drug moved from initial target identification to IND filing in just under 18 months — a process that historically takes four to six years. Phase 3 efficacy data showed a 38% improvement in lung function preservation versus standard-of-care, with a side-effect profile no worse than existing therapies.
How the model contributed
Insilico's pipeline uses three coupled generative systems: PandaOmics for target discovery, Chemistry42 for molecule design, and inClinico for trial outcome prediction. The approved molecule was the 87th candidate the system proposed for this target — but importantly, every one of those 87 was synthesized and assayed, and the model used those results to improve its next batch.
The broader pipeline
At least 27 AI-designed molecules are now in human trials globally. Big Pharma, which spent 2021–2023 oscillating between hype and dismissal, has decisively shifted: nearly every top-10 pharma company has signed a multi-year deal with an AI-first biotech in the last twelve months. Expect a steady drumbeat of similar approvals through 2027 and 2028.