In July 2026, Insilico Medicine announced it had begun a Phase III clinical trial for rentosertib, a drug for idiopathic pulmonary fibrosis whose biological target and whose actual molecular structure were both identified using generative artificial intelligence — not just AI-assisted analysis of existing candidates, but AI-driven design from the ground up. The milestone matters because rentosertib is widely regarded as the clearest real-world test case so far of whether AI-designed drugs can survive the same rigorous, multi-year clinical trial gauntlet as drugs discovered through conventional laboratory methods.
What idiopathic pulmonary fibrosis does to patients
Idiopathic pulmonary fibrosis, or IPF, is a progressive disease in which lung tissue becomes scarred and stiff over time, gradually robbing patients of the ability to breathe. It has no known single cause, existing treatments slow but do not reverse the damage, and median survival after diagnosis is commonly cited at just three to five years, comparable to many cancers. The disease is exactly the kind of complex, poorly understood condition where finding a genuinely new biological target — a fresh point in the disease process a drug can intervene on — has proven difficult using traditional discovery methods.
How AI found the target and designed the molecule
Insilico Medicine identified TNIK, a protein involved in driving pathological lung scarring, as a novel drug target using its generative AI platform, Pharma.AI, then used the same AI system to design rentosertib, a small molecule intended to inhibit TNIK’s activity. This end-to-end approach — AI proposing both what to target and what molecule should hit that target — is distinct from more common uses of AI in pharma, where machine learning typically screens or ranks candidates a human research team has already generated. The results were published in Nature Medicine, reporting on a randomized, double-blind, placebo-controlled Phase 2a trial involving 71 patients across multiple treatment centers.
What the Phase 2a results actually showed
Patients in the trial received 12 weeks of treatment at one of three rentosertib dose levels or a placebo. The 60-milligram once-daily dosing arm showed a mean improvement in forced vital capacity — a standard measure of lung function — of 98.4 milliliters at 12 weeks, a meaningful signal in a disease where lung function typically only declines over time rather than improves. The trial also reported manageable safety and tolerability across dose groups, clearing the bar needed to advance to a larger Phase III study, which is the final stage of testing generally required before a drug can be submitted for regulatory approval.
Why this case is being watched so closely across biotech
The pharmaceutical industry has invested heavily in AI drug discovery over the past decade, and the number of AI-originated drug programs entering human clinical trials has grown from roughly 3 in 2016 to more than 200 by early 2026. But most of that growth has been in program counts, not proven outcomes — until a drug clears late-stage trials and reaches patients, AI-discovery is still, in the eyes of many scientists, an unproven promise rather than a track record. Rentosertib reaching Phase III, following a peer-reviewed, statistically significant Phase 2a result, is one of the strongest pieces of evidence to date that AI-generated drug candidates can perform as well in humans as those found through decades-old discovery methods, shifting the industry conversation from speculative claims toward reviewable, published data.
The road ahead
Phase III trials are larger, longer and more expensive than Phase 2a, typically enrolling hundreds of patients across many more sites over one to several years, and the great majority of drugs that enter Phase III still fail to reach approval. Insilico Medicine has not disclosed a target date for completing the trial or filing for regulatory approval. If rentosertib succeeds, it would likely become one of the first AI-discovered drugs to reach patients, offering IPF patients a new treatment option and giving the broader AI-in-biotech industry its most concrete success story yet — a result that would ripple well beyond one disease, reshaping how much confidence regulators, investors and hospital systems place in AI-driven approaches to inventing entirely new medicines.
