Skip to content
← Back to Biology & Medicine

An AI-discovered drug for a lung disease entered final-stage trials; none is approved yet

In a 71-patient trial, lung capacity rose on the drug and fell on placebo over 12 weeks. A 320-patient final trial began in September 2026.

+98.4 mL vs −20.3 mL12-week change in lung capacity, top dose vs placebo (71 patients)

Rentosertib is Insilico Medicine’s drug candidate for idiopathic pulmonary fibrosis, a disease that scars the lungs and makes breathing steadily harder. Two things about it came from the company’s AI systems: the protein the drug acts on, and the molecule itself.

In a trial published in Nature Medicine in June 2025, 71 patients at 22 sites in China were given one of three doses or a placebo for 12 weeks. The main goal was safety, and side effects were common in every group, including placebo. On the top dose, lung capacity rose by an average of 98.4 mL, while it fell by 20.3 mL on placebo.

On 10 September 2026 Insilico gave the first dose in a Phase 3 trial, the large final stage before a company can ask a regulator for approval. It plans 320 patients at 47 centres in China and will track lung decline over a year. Insilico says the drug is still experimental and approved by no regulator.

For scale, a 2024 analysis found that molecules from AI-led pipelines passed the first stage of trials about 80–90% of the time but the second only about 40%, close to the industry’s long-run average, from a small sample.

What it means

AI-designed drugs are now being tested in people at full scale, and this one is far enough along that a real answer is coming. It is still only a candidate: a 12-week change in lung capacity in 71 patients is not proof that people live better or longer, which is what the year-long trial has to show. Nobody can buy or be prescribed it today.

Caveats

  • The earlier trial was small (17–18 patients per group) and short (12 weeks). Lung capacity was a secondary outcome with a wide margin of error (95% CI +10.9 to +185.9 mL on the top dose).
  • Side effects judged to be treatment-related were more frequent on the drug than on placebo.
  • “AI-discovered” has no agreed definition. The chemistry refinement, lab testing and trial design were conventional.
  • “No approved AI-discovered drug” rests on searches and Insilico’s own statement, not a check of regulators’ databases.