Curie Brief
Turn on cookies to sign in
Signing in saves your progress to your Curie account. We can only do that with cookies on — turn them on to continue.
News
Despite massive hype and over $40 billion in VC investment this decade, AI has yet to meaningfully transform drug development. A new peer-reviewed paper in Nature Reviews Drug Discovery calls AI's clinical impact "disappointingly limited," noting that AI-derived drugs still hit the same Phase II wall as conventional ones. The promise is real — but the results aren't there yet.
The AI-in-pharma story has been one of the decade's biggest investment narratives — but a sobering new peer-reviewed paper in Nature Reviews Drug Discovery is pumping the brakes. Researchers find that while AI has gotten good at flagging potential drug candidates, it hasn't proven those picks can survive the messy complexity of human biology. AI-derived drugs are hitting the same costly Phase II wall as conventional drugs, where failure rates are high and patient diversity makes efficacy hard to prove.
A big part of the problem? Data. Cellular datasets — the unglamorous backbone of AI drug modeling — are often too messy or incomplete for machine learning to work reliably. As biotech journalist Derek Lowe put it, "We don't know how to clean this stuff up and categorize it for real ML/AI, and honestly, we don't even know if it can be." One notable bright spot: a Moderna-Merck Phase III trial using AI to select tumor targets for cancer patients — though experts note this is personalized decision-making, not wholesale drug discovery.
By the Numbers:
Why it matters: With billions on the line and patient hopes riding high, the gap between AI's promise and its real-world clinical output is a critical reality check for investors, regulators, and the healthcare system at large.
