Anthropic shares drug-development results, citing 89% less document review time

Drug-development teams using Claude are reporting shorter document reviews and promising laboratory designs—not proven new medicines. Anthropic’s October 27, 2026 progress report describes work with six partner organizations, including an 89% reduction in human review time at Enveda. Most of the research remains preclinical, meaning the potential treatments generally have not reached testing in people.

Less time checking how drugs are made

At Enveda, manufacturing chemists had spent up to 75% of their time on paperwork, according to Anthropic. Companies need records demonstrating that each batch meets strict quality standards before conducting clinical trials. Enveda connected its manufacturing documents to Claude so the model could flag possible problems for review.

In tests covering 1,061 pages, Claude caught all 11 issues that human reviewers had identified and flagged seven more they had missed, Anthropic said. The process brought human review time below three hours, down from more than 25. That result measures a faster review process, rather than the removal of human review.

Claude also prepared first drafts of two regulatory documents. One took four hours, against about three months for Enveda’s manual process. For the other, the company’s normal human review found one minor error. These were drafting results: the comparison does not describe the time needed to secure regulatory approval.

Designs still have to survive the lab

The molecule-design studies tackle different problems. Divergence Labs wants a food-allergy treatment that detaches IgE, an antibody involved in allergic reactions, from immune cells. Its proposed drug needs two arms: one to bind IgE tightly and another to help pull it away. Standard medications instead neutralize IgE circulating in blood, Anthropic explained.

Claude turned those goals into a design plan and generated roughly 35,000 antibody designs in about a month, largely autonomously, Anthropic said. Early tests found three designs that bound IgE and 20 that dislodged it. Three of the latter appeared to dislodge more IgE than the standard medication. Combining both functions into one drug remains the next step.

Generate Biomedicines focused on CAR T-cell therapy, which uses engineered immune cells to attack cancer. The company aims to adapt the treatment from blood cancers to solid tumors. Six scientists used Claude Code to shorten part of the therapy’s genetic instructions, creating room for additional instructions suited to other cancers.

One design was about 70% shorter than the standard version. In laboratory tests, cells produced roughly 10 times as many receptors—the proteins they use to recognize cancer cells. They also killed tumor cells more efficiently and stayed active longer, Anthropic reported. Those measurements describe cell behavior in the lab, not demonstrated benefit in patients.

At BigHat Biosciences, Claude proposed changes to two antibody fragments that could become components of future drugs. The lab made and tested about 1,010 variants over roughly 10 weeks. Anthropic reported different improvements across the strongest versions:

• Target binding improved more than tenfold compared with the original fragments.
• The cancer-targeting fragment stayed in solution at a temperature 18°C higher.
• The fragment intended to keep drugs in the body longer clumped more than 100 times more slowly.

Changing the experiment—and planning the next stage

BigHat’s work also involved revisiting how success was measured. Partway through, the team suspected its standard heat-stability and clumping measurements were misleading. Claude and an Anthropic researcher reanalyzed raw data, proposed alternative measurements and suggested a revised testing procedure. BigHat scientists refined and ran it; Anthropic said the main predictions held.

The remaining partnerships test organizational and planning work. The Column Group is establishing Pleiades, a company intended to run more complex drug programs with a small team using Claude. Its first Claude skill, for assessing whether drug targets merit further work, has been evaluated on 45 target-disease pairs.

Bristol Myers Squibb is using Claude to assess early-stage clinical-trial plans and estimate their chances of success. The partners plan to expand that work in 2027. Anthropic says the collaborations have already helped it prioritize model improvements—a feedback loop in which real drug programs expose tasks Claude needs to handle better.