Phase III for a drug made with generative AI
On 10 September 2026 Insilico Medicine announced that the first patient had been dosed in GENESIS-IPF-3 (NCT07687459), a Phase III trial of rentosertib, a TNIK inhibitor for idiopathic pulmonary fibrosis whose target and molecule the company says were found and designed with generative AI. The registry gives the start as 9 September; 320 patients, 52 weeks. No Phase III result exists.
Why it matters
This is a Phase III trial of a drug the company calls created with a generative model; "the world’s first" is its own claim. The record fixes what is being tested and at what stage, not whether the drug works: nobody knows that yet.
State. Claimed: Insilico’s release of 10 September (insilico.com), read in full. The trial GENESIS-IPF-3: rentosertib (ISM001-055 / INS018_055), NCT07687459 and CTR20262475, Peking Union Medical College Hospital; Shanghai Pulmonary Hospital enrolled its first patient the same day (the release names no day); prospective, multi-centre, randomised, double-blind, placebo-controlled, parallel-group, once daily for 52 weeks; 320 participants at 47 centres in China; primary endpoint the annual rate of decline in forced vital capacity (FVC), key secondary endpoint time to first progression event. The release says plainly that the drug remains investigational and has not been approved by any regulator. The ClinicalTrials.gov registry (read on 10 October): sponsor InSilico Medicine Hong Kong Limited, Phase 3, first submitted 22 June, posted 7 July, start 9 September (actual), primary completion 30 October 2029 (estimated), no results, status recruiting. The company announced the initiation of Phase III as early as 7 July (PR Newswire); the September release is about the first patient. The company’s claims. The target TNIK was prioritised by the PandaOmics platform and the molecule "generated and optimized" by Chemistry42; the headline about "the world’s first Phase III trial of a generative AI-driven" drug and "the world’s first drug candidate discovered and developed using generative AI" are its own. On the Phase IIa study the release says "promising results, including improvements in lung function" over 12 weeks; the 7 July release gives a mean FVC improvement of +98.4 mL at 12 weeks on 60 mg once daily (the Nature Medicine paper itself was not read). The leading investigator puts the time from the start of Phase III to final approval at three to four years under favourable conditions. What the record does not claim: that the drug is effective or safe in Phase III, that Phase III results exist, or that "the world’s first" has been checked independently.