Anthropic Uses Claude to Cut Drug Review Time by 89 Percent and Double Muscle Strength in Mice

Quick Facts

• Anthropic partnered with six drug developers including Bristol Myers Squibb and Generate Biomedicines, reporting results across cancer therapy, food allergies, and clinical trial planning.
• The company acquired Mithril Medicine in July 2026, a solo-founder biotech working on two rare muscular dystrophy treatments, with mouse results showing roughly twice the grip strength in treated animals.
• Claude flagged all 11 issues human reviewers found in 1,061 pages of manufacturing records at Enveda, plus seven they missed, cutting review time by 89 percent.

Anthropic on October 27 published an early progress report on drug development with Claude, detailing work across six external partners and its first in-house drug programs. The company argues that bottlenecks in drug development are distributed across the entire process, not concentrated in any single step, and that direct involvement in real programs is the only way to find and fix them.

The report covers results ranging from molecule design to regulatory documentation and clinical trial planning. Most work is preclinical. The technical report is marked not peer reviewed.

Six Partners, Six Problems

At Generate Biomedicines, six scientists used Claude Code to redesign a component of CAR T-cell therapy. The proposed design was about 70 percent shorter than the standard version, and in lab experiments, immune cells produced receptors at roughly 10 times the normal rate. Tumor cells were killed more efficiently, and immune cells remained active for longer.

Divergence Labs, a biotech founded in 2025, is designing a food-allergy antibody that targets both free-floating IgE and IgE already bound to immune cells. Claude generated roughly 35,000 antibody designs across 22 autonomous work stretches totaling around 2,750 hours on Nvidia H100 GPUs.

At BigHat Biosciences, Claude proposed changes to two antibody fragments. BigHat made and tested about 1,010 variants in approximately 10 weeks. The best-performing versions bound more than 10 times more tightly than the originals.

Enveda connected its manufacturing records to Claude and ran the system against 1,061 pages of documents. It caught all 11 issues human reviewers had flagged, plus seven additional ones, cutting review time by 89 percent. A comparable draft had previously taken about three months by hand.

Bristol Myers Squibb is using Claude to assess early-stage drug plans and estimate the likelihood of clinical trial success. BMS early-clinical-development SVP Michael Pourdehnad said: "From a single prompt, Claude produced the initial analysis that would take days, and it did so while keeping high-sensitivity data secure." BMS and Anthropic plan to expand the work in 2027.

The Column Group, a life sciences investment firm, is using Claude to make ambitious early-stage drug discovery faster and cheaper. The firm is setting up a new company called Pleiades to run on Claude, with the goal of making riskier scientific bets more fundable.

The Mithril Acquisition

Anthropic also revealed it acquired Mithril Medicine in July 2026, though the deal was not disclosed until October 27. Mithril was a one-person company founded by Cody Tranbarger, who used Claude to coordinate 11 external labs and manage contracts, logistics, and experiments.

Mithril focused on two rare genetic disorders: LAMA2-related congenital muscular dystrophy and alpha-dystroglycanopathy. Patients with LAMA2-related muscular dystrophy are typically born with the disease, spend their lives in wheelchairs and on ventilators, and rarely survive past their mid-twenties. No approved therapy exists.

Tranbarger moved a potential LAMA2 treatment from concept to promising mouse results in under 18 months. In a six-week pilot study, mice receiving the highest antibody dose showed roughly twice the grip strength and muscle force of untreated mice.

Anthropic plans to select formal drug candidates in 2027. Human trials for LAMA2-related muscular dystrophy could begin in late 2028, with alpha-dystroglycanopathy trials potentially starting in 2029.

What This Means for the Industry

The report signals that Anthropic is moving beyond tooling and into direct drug development, competing for ground with Google and OpenAI in life sciences AI. By acquiring Mithril and launching its own programs, the company is taking on scientific and regulatory risk rather than simply licensing technology to partners.

The results to date are early and unreviewed. But the scope of the work, spanning oncology, immunology, rare disease, manufacturing, and clinical operations, suggests Anthropic is betting that AI can compress timelines across the full drug development cycle, not just one part of it.