AI chemist improves a medicinal chemistry reaction

OpenAI says a nearautonomous AI system helped improve a difficult reaction used in medicinal chemistry, working with Molecule.one’s highthroughput lab and human chemists to test ideas and analyze results. The project focused on Chan–Lam coupling, a reaction used to form carbonnitrogen bonds. The system identified primary sulfonamides as a challenging substrate class and suggested TEMPO, a mild oxidant, as a useful additive. Across two rounds of experiments, the optimized conditions raised average yield from 16.6% to 25.2% and increased the share of reactions above 30% yield. Human chemists then repeated representative reactions at bench scale and found higher yields in 11 of 14 substrate pairs. OpenAI said the work is an early example of AI helping scientists design experiments, interpret data, and develop practical results, while still relying on expert oversight and laboratory validation.