The Verification Era Begins: Insilico Doses First Patients in Phase III Rentosertib Trial

AI drug discovery has cleared plenty of Phase I gates. On July 7, 2026, it entered a harder room: Phase III.

Insilico Medicine dosed the first patients in a 320-person Phase III trial of rentosertib (ISM001-055), a molecule whose target was identified by AI, whose structure was generated by AI, and whose trial design was optimized by AI. The study is a prospective, randomized, double-blind, placebo-controlled trial across 47 centers in China, led by Professor Zuojun Xu of Peking Union Medical College Hospital. Trial identifiers include CTR20262475 and NCT07687459.

Why rentosertib matters

Rentosertib targets TNIK, an AI-discovered target, with an AI-designed small molecule. In June 2025, Insilico published Phase IIa results in Nature Medicine—the first drug with both an AI-discovered target and AI-designed molecule to show positive clinical efficacy in patients.

The primary endpoint in Phase IIa was safety; secondary endpoints included efficacy signals. Patients in the highest-dose arm (60 mg QD) saw mean forced vital capacity change of +98.4 mL (95% CI 10.9–185.9), while placebo declined by 20.3 mL. In idiopathic pulmonary fibrosis—a disease defined by irreversible lung capacity loss—that directional signal drew attention.

Phase IIa enrolled 71 patients over 12 weeks. Phase III will test durability over 52 weeks with ~320 participants.

The verification era

Roughly 175 AI-discovered programs sit in clinical trials with zero FDA approvals to date, according to industry tallies. BCG data cited in recent analyses suggests AI-discovered molecules clear Phase I at 80–90% versus a ~50% historical norm—but Phase II success falls back toward the industry-standard ~40%.

Between 2026 and 2027, an estimated 15–20 AI-discovered programs reach pivotal Phase III readouts. Rentosertib is among the most watched because the AI lineage runs through target, molecule, and trial design—not just one step.

Regulatory posture

No major regulator has suggested AI methodology alone should block approval. The credible first approval scenario remains a drug that works clinically, not a regulatory exception for AI branding.

The EU AI Act's high-risk provisions take effect August 2, 2026, and the FDA is expected to finalize AI guidance this year requiring credibility assessment plans and detailed model documentation for high-risk applications.

What success would prove

A positive Phase III would not settle every AI-drug question. It would demonstrate that an end-to-end AI pipeline can produce a therapy whose clinical outcomes survive the most expensive, statistically unforgiving stage of development.

Failure, equally informative, would reinforce that molecule design acceleration does not automatically fix target selection or clinical translation.

Sources

• AI2Work — 175 AI Drug Programs, Zero Approvals: The Phase III Reckoning (2026)
• AI2Work — Rentosertib: The AI-Designed Drug That Rewrote Drug Discovery (July 22, 2026)
• Drug Target Review — AI in drug discovery: predictions for 2026 (2026)