AI Drug Discovery Reaches a Reckoning: The First Clinical Trials Tell the Truth
By Sanna the Weaver • Sat Feb 21 2026 • Health
The promise of AI-driven drug discovery has been extraordinary. Companies including Recursion Pharmaceuticals, Exscientia, Insilico Medicine, and Isomorphic Labs (the drug discovery spinoff of DeepMind) have spent years arguing that machine learning can identify drug candidates faster, cheaper, and more successfully than traditional discovery methods. In 2026, the clinical data necessary to test that argument is beginning to arrive — and the results are more complicated than the hype suggested. What AI Drug Discovery Actually Does AI drug discovery is not a single technology but a collection of machine learning applications applied to different stages of the drug development pipeline. Generative AI can propose novel molecular structures with desired properties. Predictive models can estimate a molecule's likelihood of binding to a disease target, its metabolic stability, its toxicity, and its potential off-target effects. AlphaFold (DeepMind) has essentially solved protein structure prediction — knowing the 3D shape of a target protein dramatically accelerates the design of molecules that fit it. Together, these tools can dramatically compress the timeline from target identification to preclinical candidate selection, which traditionally takes five to seven years. The Clinical Reckoning The bottleneck has always been clinical trials — the human testing that biology ultimately requires and that no AI can shortcut. Insilico Medicine's ISM001-055, an AI-designed drug for idiopathic pulmonary fibrosis, completed Phase 2 trials in 2024 and showed promising results. Recursion Pharmaceuticals' REC-2282, targeting a rare brain condition, is in Phase 2. Exscientia's first AI-designed drug candidate failed in Phase 1 — a reminder that AI can optimize for the variables it can model but cannot predict the full complexity of human biology. The early scorecard shows AI-designed drugs performing roughly comparably to traditionally discovered drugs at the same stage — which means most will fail, because most drugs fail regardless of how they were discovered. "AI has made us much faster at getting to clinical trials. It has not yet made us more likely to succeed in them." — Chief Scientific Officer, Recursion Pharmaceuticals, March 2026 Where AI Is Already Delivering The clearest near-term value of AI in drug development may be in repurposing — identifying existing approved drugs that could be effective against new targets. This bypasses the years-long safety evaluation process since the drug's toxicity profile is already known. AI systems have identified several clinically validated repurposing opportunities in oncology and rare diseases, some of which are moving quickly through trials. Meanwhile, BioNTech and Moderna are using AI to accelerate the design of mRNA cancer vaccines personalized to individual tumor mutation profiles — with the first patients receiving such vaccines in expanded trials in 2026. Here, the AI advantage in speed is unambiguous and the preliminary clinical signals are encouraging.