Fold Commons

ClinicalAtlas

A transparency record of AI and machine learning in clinical trials — the AI-designed drug candidates that reached human testing, the AI/ML diagnostic devices cleared as trial endpoints, and the AI-driven repurposing hypotheses carried into trials. Each milestone links to its primary source — a company announcement, a ClinicalTrials.gov record, an FDA authorization, or a peer-reviewed paper. Nothing is installed; nothing about you is collected.

This is the tool — a curated landmark set, not a systematic index; it is drawn only from free public sources, so the picture is deliberately partial — see the coverage gauge below. ClinicalAtlas describes what was announced or published; it never asserts causation, never rates whether an AI use was effective or appropriate, and never speculates beyond the cited record. Editorial is CC BY 4.0; each milestone carries its own citation.

Honest coverage: ≈20% of the landmark record

Built from free public sources, this tracker captures roughly 20% of publicly-verifiable AI-in-clinical-trials milestones. We show what we can source and name what we can’t, so you can calibrate your conclusions. As of 2026-09-17.

What’s covered

  • AI-designed or AI-discovered drug candidates that reached human clinical trials, with a company/press or registry trail
  • Regulatory-clearance milestones for AI/ML medical devices used as trial endpoints or screening tools (FDA)
  • Landmark AI-repurposed drug candidates entering trials
  • Notable AI-in-trial-operations milestones with a primary source

Known gaps

  • Most industry AI-discovery pipelines disclose only at press-release granularity and never publish an NCT id — those are omitted here rather than guessed
  • AI used inside trial operations (site selection, recruitment, imaging endpoint adjudication) is rarely disclosed in registry text and is under-counted
  • EU CTIS and non-US registry entries are under-represented relative to ClinicalTrials.gov
  • Failed, paused, or terminated AI-associated trials are under-represented (survivorship bias toward announced milestones)
  • This is a curated landmark set, not a systematic index of every AI-associated trial
Area
Phase
Status

10 of 10 milestones

  1. Review / meta-analysis Completed 2022

    First cohort of AI-discovered molecules in trials — independent review benchmark

    BCG / Wellcome (Jayatunga et al.)

    An independent analysis in Nature Reviews Drug Discovery examined the first AI-discovered molecules to reach clinical trials, reporting early Phase I success rates for AI-native biotech candidates and providing the coverage-denominator context for how many AI-discovered drugs had actually entered trials by 2022 — the honest baseline this tracker calibrates against.

    AI in trial operations Review / meta-analysis Completed

    Primary paper: AI in small-molecule drug discovery: a coming wave? (Nature Reviews Drug Discovery, 2022) · doi:10.1038/d41573-022-00113-2

    Sources (1)
  2. Phase 1 Active 2022

    AI-discovered 3CL protease inhibitor programs entering trials (post-COVID pipeline)

    Insilico Medicine

    Following its IPF program, Insilico Medicine advanced additional generative-AI-discovered candidates into first-in-human studies across oncology and other indications, expanding the count of AI-discovered small molecules in early clinical development. These programs are disclosed at company/press granularity; specific NCT ids are omitted here where not independently confirmed.

    AI-discovered / AI-designed drug Phase 1 Active

    Sources (2)
  3. FDA cleared Completed 2021

    GI Genius — first FDA-cleared AI polyp-detection device (computer-aided colonoscopy)

    Medtronic / Cosmo Pharmaceuticals / FDA

    GI Genius was authorized by the FDA in 2021 as the first computer-aided detection device using AI to identify colorectal polyps in real time during colonoscopy. Its randomized supporting evidence (adenoma detection rate as a trial endpoint) illustrates AI as an investigational device with a prospective randomized endpoint.

    AI diagnostic device FDA cleared Completed

    Primary paper: Efficacy of Real-Time Computer-Aided Detection of Colorectal Neoplasia in a Randomized Trial (Gastroenterology, 2020) · doi:10.1053/j.gastro.2020.04.062

    Sources (2)
  4. Phase 2 Active 2021

    INS018_055 (rentosertib) — AI-discovered TNIK inhibitor for idiopathic pulmonary fibrosis

    Insilico Medicine

    A small-molecule TNIK inhibitor for idiopathic pulmonary fibrosis (IPF) — presented by Insilico as the first drug with both an AI-discovered novel target and an AI-generated molecular structure to enter clinical trials. First-in-human Phase I dosing began in 2021/2022; positive Phase IIa topline results were reported by Insilico in November 2024, and the peer-reviewed Phase IIa results were published in Nature Medicine in 2025. Later assigned the name rentosertib.

    AI-discovered / AI-designed drug Phase 2 Active

    Primary paper: A generative AI-discovered TNIK inhibitor for idiopathic pulmonary fibrosis: a Phase 2a trial (Nature Medicine, 2025) · doi:10.1038/s41591-025-03743-2

    Sources (2)
  5. FDA cleared Completed 2021

    Paige Prostate — first FDA-authorized AI in digital pathology

    Paige.AI / FDA

    Paige Prostate received FDA De Novo authorization in September 2021 as the first AI-based software to aid pathologists in detecting areas suspicious for cancer in prostate biopsies — the first authorized AI application in digital pathology, a category increasingly used to define imaging/pathology endpoints in oncology trials.

    Digital pathology AI FDA cleared Completed

    Sources (1)
  6. Phase 2 Active 2021

    REC-994 (tempol) — AI-platform candidate for cerebral cavernous malformation in trials

    Recursion Pharmaceuticals

    Recursion advanced REC-994, a superoxide scavenger identified/prioritized through its image-based phenomics discovery platform, into clinical development for symptomatic cerebral cavernous malformation (CCM). The randomized Phase 2 SYCAMORE trial (NCT05085561) was a leading example of a candidate from an AI/ML-driven phenotypic drug-discovery platform reaching controlled trials.

    AI-discovered / AI-designed drug Phase 2 Active

    Sources (2)
  7. Phase 3 Completed 2020

    Baricitinib for COVID-19 — AI-driven drug repurposing hypothesis validated in trials

    BenevolentAI / Eli Lilly

    BenevolentAI used its knowledge-graph AI to hypothesize in early 2020 that the JAK inhibitor baricitinib could reduce SARS-CoV-2 infectivity and inflammation. The hypothesis was published in The Lancet, tested in the ACTT-2 randomized trial, and baricitinib subsequently received FDA authorization/approval for hospitalized COVID-19 patients — a prominent example of an AI-generated repurposing hypothesis carried through to trials.

    AI-driven repurposing Phase 3 Completed

    Primary paper: Baricitinib as potential treatment for 2019-nCoV acute respiratory disease (The Lancet, 2020) · doi:10.1016/S0140-6736(20)30304-4

    Sources (2)
  8. Phase 1 Discontinued 2020

    DSP-1181 — first AI-designed drug molecule to enter a human clinical trial

    Exscientia / Sumitomo Dainippon Pharma

    A long-acting serotonin 5-HT1A receptor agonist for obsessive-compulsive disorder, created with Exscientia's Centaur Chemist AI platform. Announced in January 2020 as the first AI-designed molecule to reach a Phase I human trial. The Phase I study in Japan was later discontinued; the milestone stands as the first-of-its-kind entry.

    AI-discovered / AI-designed drug Phase 1 Discontinued

    Sources (2)
  9. FDA cleared Completed 2018

    IDx-DR — first FDA-authorized autonomous AI diagnostic (diabetic retinopathy)

    Digital Diagnostics (IDx) / FDA

    IDx-DR received FDA De Novo authorization in April 2018 as the first medical device authorized to provide a screening decision (for more-than-mild diabetic retinopathy) without clinician interpretation. Its pivotal prospective trial (published in npj Digital Medicine) is a template for how an autonomous AI diagnostic is validated as a study endpoint.

    AI diagnostic device FDA cleared Completed

    Primary paper: Pivotal trial of an autonomous AI-based diagnostic system for detection of diabetic retinopathy (npj Digital Medicine, 2018) · doi:10.1038/s41746-018-0040-6

    Sources (2)
  10. FDA cleared Completed 2018

    Viz.ai LVO — first FDA-cleared AI triage/notification device (stroke)

    Viz.ai / FDA

    Viz.ai's ContaCT (LVO) received FDA De Novo clearance in February 2018 as the first computer-aided triage and notification software using AI to analyze CT scans for suspected large-vessel-occlusion stroke and alert specialists — establishing the AI clinical-decision-support / triage category that later trials use as an operational intervention.

    AI diagnostic device FDA cleared Completed

    Sources (1)

Methods & limits