A public ledger — updated when the number changes Last checked · Nov 21, 2026

Number of diseases cured by AI

0
Diseases cured Verified · FDA / EMA approvals Est. forever ago

Zero. Not “almost one.” Not “one, if you count a scan.” No disease has been cured by a cure that AI discovered — no AI-discovered drug has even been approved yet.

What would count? ↓

Snapshot as of Nov 21, 2026 — checked against FDA approvals through October 2026. This page does not poll live.

1,452 days since ChatGPT launched 1,452 of those days at zero $6.93B AI drug-discovery market in 2025 — and counting
0
Diseases cured by an AI-discovered treatment. The whole point of this page.
0
AI-discovered drugs approved by the FDA or EMA. None, as of Oct 2026.
117
AI-enabled candidates that had entered human trials by July 2025, from 63 companies.
1
Reached Phase III. Rentosertib for lung fibrosis — still investigational, not a cure.

What would actually move this to 1?

The bar is not mysterious. It is the same bar every other drug has to clear. AI does not get a discount for being exciting in a keynote.

Counts All three have to be true

  • ✓AI found it. The target and the molecule were both chosen or designed by AI — not just “AI was used somewhere in the building.”
  • ✓Regulators approved it. It finished Phase III and was approved by the FDA, EMA, or an equivalent regulator for that disease.
  • ✓It cures. Patients are cured — disease eliminated or durably reversed — not just symptoms managed a little better for a little longer.

Doesn’t count Impressive ≠ a cure

  • ×Reading a scan. 1,614 AI devices are FDA-authorized, mostly for radiology. Detecting is not curing.
  • ×Predicting a protein. AlphaFold is a genuine breakthrough. A predicted shape is not a treated patient.
  • ×Still in a trial. Phase I, II, or III is a promise with paperwork. This counter only moves on approval + cure.
  • ×Helping in the lab. Faster screening, better papers, cheaper discovery — valuable, and still not a cure.

Closest to breaking the zero

If the number ever changes, it will almost certainly start here. None of these is there yet — here is exactly how far each one is.

01

Rentosertib — Insilico Medicine · idiopathic pulmonary fibrosis

The furthest along by a clear margin, and the first drug where AI picked both the target (TNIK) and the molecule. Phase IIa in Nature Medicine (71 patients, 12 weeks) showed a +98.4 mL lung-function signal at the top dose. Phase III started in 2026 — 320 patients, 52 weeks. IPF has no cure today; even success here would first be a treatment, not yet a cure.

Phase III · started Sep 9, 2026First patient dosedReadout still ahead
Distance to 1
Phase III → ?
02

Zasocitinib — Takeda / Nimbus · psoriasis

The nearest FDA decision: the FDA accepted its application with a target date in Q1 2027 after it beat an approved rival in Phase III. The catch for this counter: its roots are physics-based modelling from around 2016 against an already-known target, and psoriasis treatment is management, not a cure.

NDA under Priority ReviewDecision due Q1 2027Treats, does not cure
Distance to 1
Filed, not cured
03

Isomorphic Labs — Alphabet / DeepMind spinout

The loudest promise and the thinnest clinical record. Raised $2.1B in May 2026 on the strength of AlphaFold. As of November 2026, no Isomorphic molecule had been dosed in a human — first-in-human was targeted for the end of 2026.

$2.1B Series B · May 2026No human dosing yetNo named candidate disclosed
Distance to 1
Pre-clinical