Anthropic’s logo. Image: Anthropic
Anthropic is no longer only selling AI to drug companies. It said on Tuesday that it has bought Mithril Medicine, a start-up with a single full-time employee that has been developing treatments for two rare childhood forms of muscular dystrophy, and that Claude does much of the scientific and operational work on these, its first in-house drug programs.
It also gave the first results from six drug developers its life sciences team has worked with for months: BigHat Biosciences, Bristol Myers Squibb, Divergence Labs, Enveda, Generate Biomedicines and The Column Group. Almost all of it is preclinical, with no drug yet tested in people, and the figures come from the companies’ own reports, not independent review.
One founder, 11 outside labs and Claude #
Mithril’s founder, Cody Tranbarger, ran the company with part-time consultants and third-party labs before Anthropic bought it in July. According to Anthropic, he used Claude to organise contracts, logistics and experiments with 11 outside labs, and took an antibody treatment for LAMA2-related congenital muscular dystrophy from an idea to promising results in mice in under 18 months.
The disease affects an estimated 4,000 to 5,000 people in the US and Europe, Mithril’s technical report says. Most children with it never walk unaided, few live to 30, and there is no approved treatment. Mithril’s antibodies are built to clamp weakened muscle fibres back onto the protein sheet that holds them. In mice bred to have a severe form of the disease, the highest dose roughly doubled grip strength and muscle force after six weeks, and one of eight treated mice died, against four of eight untreated ones. They were still weaker than healthy mice.
Anthropic calls the set-up “Pipeline-in-a-Person”: the scientist decides, while Claude keeps the records and coordinates the labs. Tranbarger estimates that turning a new drug idea into a detailed plan took two to three days with Claude, against one to two months without it. Mithril aims to name formal drug candidates in 2027 and, if they hold up, to start human trials in LAMA2-CMD as soon as late 2028 and in α-dystroglycanopathy as soon as 2029. Anthropic calls those dates targets, not commitments.
A shorter cancer receptor and 35,000 allergy antibodies #
Generate Biomedicines is trying to make CAR T-cell therapy, so far used mainly for blood cancers, work against solid tumours. Its scientists used agents built with Claude Code to redesign part of the receptor, and one design about 70% shorter than the standard version led to roughly 10 times as many receptors on human T-cells in the lab and killed tumour cells more efficiently, Generate’s preprint reports.
For Divergence Labs, which wants a food allergy drug that pulls the allergy molecule IgE off immune cells, Claude generated about 35,000 antibody designs in roughly a month, largely on its own. Early tests found three that bind IgE and 20 that dislodge it, three of which appeared to dislodge it more than the standard medicine, Divergence’s preprint says.
Paperwork cut from months to hours #
At Enveda, where manufacturing chemists spent up to 75% of their time on documents, Claude checked 1,061 pages of manufacturing records, caught all 11 issues human reviewers had found and flagged seven they had missed, cutting review time by 89%, Enveda reports. It also wrote a first draft of a regulatory document in four hours, against about three months by hand. BigHat Biosciences made and tested about 1,010 antibody variants that Claude proposed in about 10 weeks, and the best bound more than 10 times more tightly than the originals, BigHat’s report says.
Bristol Myers Squibb, the one big drugmaker in the group, uses Claude to assess trial plans for an early-stage drug and their odds of success. Its head of early clinical development for haematology, oncology and cell therapy said:
From a single prompt, Claude generated a robust initial analysis that helped inform our assessment of different development options, which would otherwise take days to complete. Michael Pourdehnad, Bristol Myers Squibb
Early days, and mostly Anthropic’s word #
Anthropic says it will scale the work up “substantially” and extend Claude into clinical trials as programs advance, with targets including food allergies, rare childhood diseases and several cancers. It follows Claude turning up a new CRISPR-like system and designing antivenom proteins that two labs say work, while OpenAI pushes the same way with GPT-Rosalind. None of the drug results has been peer reviewed yet, and many treatments that work in mice fail in people.
Why it matters #
AI labs have mostly sold tools to drugmakers; owning drug programs puts Anthropic in the business of making medicines itself, a day after it filed for its stock market listing. If one scientist with Claude can really run several programs, rare diseases too small to pay for a full team could get treatments, but for now that rests on mouse data and company reports.
Sources: Anthropic (October 27, 2026); Mithril Medicine technical report; Generate Biomedicines preprint; Divergence Labs preprint; BigHat Biosciences report; Enveda report