For years, “AI in drug discovery” mostly meant software helping humans sift through candidates faster. A drug called INS018_055 is different: artificial intelligence both identified its disease target and designed the molecule. In April 2026, it posted early human efficacy data — one of the clearest signs yet that AI is moving from the screening stage into the medicine cabinet.
What the AI did
Developed by Insilico Medicine to treat idiopathic pulmonary fibrosis — a serious, progressive scarring of the lungs — INS018_055 was generated end to end by the company’s AI platform, which pinpointed the biological target and then designed a novel compound to hit it. Insilico took the drug from concept to a Phase 1 trial in about 30 months, a fraction of the usual timeline.
The results
On April 18, 2026, the company reported Phase IIa data showing a dose-dependent efficacy signal — the kind of clean, dose-related response researchers hope to see. Insilico’s CEO called it a critical milestone, noting they hadn’t expected such a clear signal after so short a dosing period. Early, but genuinely encouraging.
The bigger picture
The momentum is industry-wide. More than 200 AI-discovered drugs are now in clinical trials, and in March 2026 Eli Lilly signed a collaboration with Insilico worth up to $2.75 billion — the largest AI-era pharma deal yet. Google DeepMind’s Isomorphic Labs, meanwhile, has said it aims to put its first AI-designed candidate into Phase 1 by the end of 2026.
The caveat that matters
Here’s the sober counterweight: despite 200-plus candidates in trials, exactly zero AI-designed drugs have won FDA approval. Phase IIa is early, most drugs that look promising here still fail later, and AI doesn’t change the hard biology of large, lengthy human trials. The technology compresses discovery — not the years of proof that follow.
Why it matters
Designing drugs is one of the slowest, most expensive endeavors in science, and the bottleneck has always been finding the right molecule. If AI can reliably shorten that front end — as INS018_055’s timeline and early signal suggest — it could mean more shots on goal for diseases that badly need them. The first AI-designed medicine isn’t approved yet, but it’s now far enough along to take very seriously.
