For as long as pharmaceutical companies have been developing drugs, the process has followed a roughly similar arc: identify a disease target, screen millions of compounds for biological activity, optimise the most promising candidates, run years of animal trials, then navigate the long and expensive gauntlet of human clinical trials. The whole process, from target identification to approval, typically takes 10 to 15 years and costs more than a billion dollars.
In March 2026, the FDA approved a drug called INX-750 for the treatment of a rare blood cancer — a milestone that attracted attention not because of the disease it treats, but because of how it was found. INX-750 was identified by an AI system, optimised by an AI system, and had its toxicity profile predicted by an AI system. The discovery-to-trial timeline was 18 months.
“We didn’t design this molecule in the traditional sense,” said Dr. Andrew Hopkins, chief scientific officer at Exscientia, the Oxford-based company that developed the drug. “The AI identified the chemical space we should be exploring and then navigated that space in ways our chemists would not have prioritised. It found something genuinely unexpected.”
The approval is significant in part because of what it reveals about the regulatory process. The FDA reviewed INX-750 on the same basis as any other drug — safety and efficacy data from clinical trials — without establishing special standards for AI-discovered compounds. That pragmatic approach has been welcomed by industry but raises questions about long-term oversight.
The drug itself treats fewer than 10,000 patients in the US annually. But the approval is being read by the pharmaceutical industry as a proof of concept for something much larger. Several major drug companies have announced AI-first drug discovery programmes in the past year, and the race to compress the decade-long drug discovery process into months is now genuinely underway.
Leave a Reply