Home / AI

An AI-Discovered Pulmonary Fibrosis Drug Enters Phase III, the Field's Biggest Test Yet

Insilico Medicine's rentosertib is in a 320-patient pivotal trial, but no AI-designed drug has yet won FDA approval.

AIJul 7, 20264 min read2 diagrams

Key takeaways

  • Insilico registered a Phase III trial of rentosertib for idiopathic pulmonary fibrosis in July 2026: about 320 patients over 52 weeks.
  • Trackers count more than 173 AI-designed programs in clinical development, up from roughly two dozen in late 2023.
  • As of July 2026, no AI-discovered drug has full FDA approval; the 'first' label is a company claim.

Insilico Medicine has launched a Phase III trial of rentosertib, an oral drug for idiopathic pulmonary fibrosis whose target and molecular structure were both identified with generative AI. It is the clearest test yet of whether machine-designed medicines can survive the stage where most drug candidates fail.

What happened

In early July 2026, the trial, named GENESIS-IPF-3, appeared in registries and the company later said it had dosed the first patient. The design is a prospective, multi-center, randomized, double-blind, placebo-controlled study with once-daily dosing over 52 weeks, enrolling approximately 320 patients. Rentosertib inhibits TNIK, a protein implicated in scarring of lung tissue, and earlier Phase IIa results published in Nature Medicine showed dose-dependent changes in forced vital capacity, the standard measure of lung function in this disease. The FDA granted the drug orphan designation for IPF back in February 2023.

Insilico describes rentosertib as the first drug for which both a novel target and the molecule itself were found using generative AI to reach pivotal-stage development. That is a company-asserted milestone, and it depends on definitions. Generate:Biomedicines, for instance, began a Phase III study of an AI-engineered antibody, GB-0895, on December 3, 2025. Which one counts as first depends on how strictly "AI-designed" is defined.

Bar chart of AI-designed drug programs in clinical development, late 2023 versus early 2026 AI-designed programs in clinical development about 24 173+ Late 2023 Early 2026 Tracker counts vary by definition; late-2023 figure is approximate
Growth in AI-designed programs in the clinic, per one tracker; counts differ across sources. Source: IntuitionLabs.

The hard part is still ahead

Discovery is only the first stretch of a long pipeline. Pivotal trials are where efficacy has to be shown against placebo in hundreds of patients, and historically a majority of candidates entering Phase III or its equivalent fail to reach approval. AI shortens the early search for targets and molecules, and Insilico has argued it can cut the time and cost of getting to the clinic. It does not change the biology that decides whether a drug works in patients.

The field has seen setbacks. BenevolentAI's BEN-2293 missed its efficacy endpoints in 2023, a reminder that an AI-nominated candidate faces the same odds at the clinic door as any other. What has changed is volume: more than 173 AI-designed programs are now tracked, versus a couple of dozen two and a half years ago. With that many shots, the field will produce both its first approval and its first high-profile Phase III failure within a few years.

Flow diagram of the drug development pipeline showing where rentosertib sits Target and molecule (AI) Preclinical Orphan status 2023 Phase IIa Nature Medicine Phase III 320 pts, 52 wks FDA AI compresses the first box; the later boxes still depend on human biology. Illustrative.
Where rentosertib sits in the development pipeline. Stages simplified; Phase III design per company announcements. Illustrative.

AI in medicine is further along elsewhere

Drug discovery is the glamorous end of healthcare AI, but the larger footprint is in devices. The FDA's public list of AI-enabled medical devices had reached 1,524 entries by March 30, 2026, with about 1,163 of them, roughly three-quarters, in radiology. In the first quarter of 2026 the agency authorized 92 such devices, 28% more than in the previous quarter. The FDA itself cautions that its list is built mainly from AI-related terms in device summaries and is not comprehensive, so the true count is likely higher.

The contrast is useful. Devices that read scans are cleared on narrower evidence and have a relatively fast path to market. A drug has to prove it helps patients live better or longer, which takes years and hundreds of patients. The two tracks will test different claims about AI in medicine.

Why it matters

If rentosertib succeeds, it validates a thesis that has attracted billions of dollars of investment: that machine learning can find drug targets humans overlooked. If it fails, it will not disprove AI drug discovery, but it will reset expectations. IPF is a serious, progressive disease with limited treatments, so a positive result would matter to patients regardless of how the molecule was found.

What to watch

  • Enrollment progress and any interim safety signals in GENESIS-IPF-3.
  • GB-0895's Phase III path, the other candidate for an early AI-linked approval.
  • Whether the FDA issues specific guidance on AI-derived drug candidates.
  • Quarterly additions to the FDA's AI device list, a measure of clinical adoption.

Sources

  1. Insilico Medicine Launches Phase III Trial for AI-Developed Drug — Bio-IT World
  2. AI-Designed Drug Rentosertib Advances into Phase III Trial for Idiopathic Pulmonary Fibrosis — GeneOnline
  3. AI-discovered drugs in clinical trials (2026) — IntuitionLabs
  4. List of Artificial Intelligence-Enabled Medical Devices — U.S. FDA

More in AI