Insilico Medicine said on Tuesday that it has begun a Phase III clinical trial for rentosertib, an oral small molecule for idiopathic pulmonary fibrosis, a serious and progressive lung scarring disease with few good treatment options. What makes the announcement more than routine pharmaceutical news is how the drug came to exist. Rentosertib, previously known as ISM001-055, was both discovered and designed using Insilico's Pharma.AI platform, which puts it among the furthest advanced examples of a medicine where artificial intelligence did the core early science rather than assisting a human led program at the margins.
It helps to be specific about what the AI actually did, because that is where the real claim lives. Insilico used its models to help identify the biological target, an enzyme called TNIK that is implicated in fibrosis, and then to generate and refine the molecule aimed at it. The pitch that the company and the broader field have been making for years is that this kind of generative approach can compress the slow, expensive early stages of drug discovery, the part where scientists hunt for a promising target and then search enormous chemical spaces for something that might work. Rentosertib is that pitch made concrete, a specific molecule that came out of that pipeline and is now deep into human testing.
The results so far are why it earned a Phase III rather than stalling. In the earlier Phase IIa study, called GENESIS-IPF and published in the journal Nature Medicine, the drug showed manageable safety and tolerability, and the arm receiving 60 mg once a day showed a mean improvement in forced vital capacity, a standard measure of lung function, of 98.4 mL at 12 weeks. In a disease defined by lung function steadily declining, a measured improvement is a meaningful signal. Insilico is now advancing the oral version into Phase III while also moving an inhaled formulation of the same molecule forward in parallel.
Now the part that most of the excited coverage tends to skip. Starting a Phase III trial is a milestone, not a victory. Phase III is the large, costly, and statistically decisive study that regulators lean on most, and a great many drugs that look promising in mid stage testing still fail once they reach it. Just as important, the fact that an AI designed the molecule does not tilt those odds, because the trial tests biology, not software. The model's contribution was upstream, in finding and shaping the candidate, and from here rentosertib has to succeed or fail on exactly the same clinical evidence as any drug designed the traditional way.
With that caveat in place, the milestone still matters, and here is the honest reason. AI drug discovery has spent years being long on promise and short on late stage proof, full of impressive early screens and partnerships but thin on candidates that reached the trials that actually decide whether a medicine gets approved. A fully AI discovered drug entering Phase III moves the conversation from what the technology might do to what one of its outputs is now being measured on, in the most rigorous setting medicine has. It is not proof that AI can deliver new drugs, and it should not be sold as such. It is the field finally putting the idea where it can be tested for real, which is exactly where a claim this big needs to end up.