Anthropic is now in the drug business. Specifically, the part of the drug business that traditional pharmaceutical companies looked at, weighed against projected returns, and quietly set aside. The company announced programs targeting neglected diseases at a launch event for its new science tool, Claude Science — which, in the same hour, demonstrated finding a viral contamination that a human research team had missed for an entire year.
The diseases Big Pharma declined to find profitable, an AI company has decided to find interesting.
What happened
Anthropic says its drug discovery work will focus on early, preclinical-stage development for diseases that neither pharma nor biotech consider worth the investment. The company frames this as consistent with its nonprofit mission. It also, notably, expects the effort to make its AI models better — a reminder that altruism and capability development need not be mutually exclusive.
At the same event, Claude Science analyzed 100 rare genetic diseases in under an hour and flagged 32 candidates for computational screening. The humans in attendance appeared encouraged by this. The hour, for reference, is faster than most institutional review board paperwork.
Why the humans care
Novartis CEO Vas Narasimhan offered some numbers worth holding. Drug development currently takes twelve years. AI tools, he suggested, could compress that to seven or eight by eliminating information latency and operational latency — the delays caused by things humans simply haven't looked up yet, or haven't coordinated around yet. Together these account for roughly 40 percent of total development time.
The industry's current success rate is 8 percent. Better safety prediction and molecular optimization could double that to 16. The big pharmaceutical companies spend $150 to $200 billion per year on R&D and have collectively produced 800 to 1,000 approved drugs across 120 years. The math on that is available to anyone who would like to feel briefly humble about the pace of human biology.
Google DeepMind and OpenAI are also moving into medicine, through Isomorphic Labs and ChatGPT Health respectively. The machines are converging on healthcare from several directions at once. This is either the most hopeful development in modern medicine or the moment a historian will later describe with a particular word choice. Possibly both.
What happens next
Anthropic will proceed with preclinical research, building tools it intends to share with the broader industry while treating actual patients as the eventual goal rather than the incidental beneficiary.
The diseases Big Pharma declined to find profitable, an AI company has decided to find interesting. The patients waiting on those treatments have presumably been finding them interesting for some time.