OpenAI released a report in April 2026 titled Reigniting the Discovery Engine for Tomorrow’s Cures. The paper explains how AI can address slowdowns in biomedical progress and is written to be clear for readers new to the topic.
Life sciences work produced antibiotics that saved hundreds of millions of lives and vaccines that eradicated smallpox while adding more than 30 years to global life expectancy. The report details how that pace has since declined. Drug development now costs about 2.6 billion dollars per approved treatment. Clinical trials average 7.5 years with only 7.9 percent of candidates that enter initial testing reaching approval.
The report describes Eroom’s Law as the pattern in which the number of new drugs approved per billion dollars of research spending has fallen roughly 80 fold since 1950 despite major gains in tools such as DNA sequencing. It also examines what it terms the Great Endarkenment. Scientific knowledge has grown so large that researchers must train longer and focus on narrower specialties. Fields once connected have split into over 20 separate areas in biomedicine alone each with distinct terms and methods. This makes it difficult to combine insights across gene cell tissue and organism scales.
AI offers a practical response. The report states that about 200000 people worldwide use ChatGPT each week for graduate level life sciences tasks. Around 75 percent of this use involves reviewing and synthesizing scientific literature across fields while 25 percent supports data analysis. Top countries for this activity include the United States India Great Britain South Korea and Germany.
Concrete results have already appeared. AlphaFold solved a decades old challenge in predicting protein shapes and provided structures for more than 200 million proteins. Groups using it submitted experimental structures at 45 to 49 percent higher rates with greater novelty. AI has also aided gene editing tools based on CRISPR with models designing variants that outperform natural ones. Self driving labs that pair AI with robotics have compressed testing cycles from months to days. One effort identified a fibrosis drug candidate in 18 months and reached early clinical testing in under 30 months total.
The report profiles several use cases. A patient with returning rare cancer integrated his imaging blood tests and sequencing data with ChatGPT to generate questions for specialists and stay ahead of disease progression. This led to a new company that found treatment options in all 15 tumors it analyzed in its first eight months. A family received genome sequencing results that an AI tool helped translate into understandable actions leading to confirmed diagnosis and community support. A doctor used the model to identify an existing drug that resolved a severe food allergy in multiple patients and now feeds lab results back into it to refine patient profiles for related conditions.
Small teams at biotech companies employ AI as a shared reference point to coordinate across clinical toxicology and discovery roles without repeated specialist handoffs. One such firm advanced a diabetes treatment program toward 2026 trials at lower than typical cost. The report concludes that AI reduces the coordination barriers created by specialization. Policy sections recommend treating advanced AI as shared research infrastructure opening machine readable datasets under privacy rules investing in connected lab and compute capacity and expanding training at the biology and computation intersection.
The report presents these applications as active workflows producing results today rather than future promises. It positions AI as a translator that lets teams see more of the full picture and accelerate the discovery process that has stalled under accumulated knowledge.