Anthropic Reports Claude Cut Enveda's Document Review Time 89 Percent, but No Drug in Its Drug Program Has Reached a Human

Anthropic published a case-study update on Oct. 27 saying Claude has helped six drug-development partners over several months, including an 89 percent cut in human review time at Enveda. The post, from Anthropic's Science team, says the work is early and mostly preclinical, and that no drug has reached human testing.

The company also said it acquired Mithril Medicine, a one-person company developing treatments for two rare forms of muscular dystrophy, and plans to scale its own drug-discovery effort. Every figure below comes from Anthropic's post, and none has been checked by an outside party.

The partner results, as Anthropic reports them

Partner	Task	Reported result
Enveda	Manufacturing paperwork review	11 of 11 human-found issues caught, 7 more flagged; review time down 89%
BigHat Biosciences	Antibody lead optimisation	About 1,010 variants tested in about 10 weeks; best binds more than 10x tighter
Divergence Labs	Food-allergy antibody design	About 35,000 designs in about a month; 3 of 20 beat standard medication
Generate Biomedicines	CAR T-cell component redesign	One design about 70% shorter; roughly 10x more receptors

At Enveda, manufacturing chemists previously spent up to 75 percent of their time on paperwork, Anthropic said. Claude reviewed 1,061 pages and caught all 11 issues human reviewers found, then flagged seven they had missed. Human review time fell from more than 25 hours to under three. Claude also drafted two regulatory documents; one took four hours against about three months by hand, and the other had one minor error caught in standard review.

At BigHat, the team found that the standard heat-stability and clumping metrics were misleading. Claude and an unnamed Anthropic researcher reanalysed the data and proposed new metrics and a protocol change. Against the originals, the best variants bound more than 10 times tighter, one stayed in solution 18 degrees Celsius higher, and the other clumped more than 100 times more slowly.

1. Before Claude (more than)25 hours
2. With Claude (under)3 hours

Source: Anthropic Science team post, Oct. 27, 2026; Anthropic-reported, accessed 2026-10-28

What the early hits do not show

Divergence Labs targets food allergies by detaching IgE, the antibody involved in allergic reactions, from immune cells. Claude generated about 35,000 antibody designs largely autonomously in about a month, Anthropic said. Early tests found three designs that bind IgE and 20 that could dislodge it, and three of the 20 dislodged IgE more than the standard medication. The drug combining both arms has not been built.

At Generate Biomedicines, six scientists used Claude Code to propose redesigns of a CAR T-cell component aimed at extending the therapy to solid tumors. One design was about 70 percent shorter than standard and, in lab tests, allowed roughly 10 times more receptors, killed tumor cells more efficiently and kept immune cells active longer.

The Pleiades skill, which judges whether drug targets are worth pursuing, was evaluated on 45 target-disease pairs. The Column Group is establishing Pleiades as a new company. Anthropic lists the BMS and Mithril timelines as projections.

Bristol Myers Squibb and the 2028 timeline

Bristol Myers Squibb uses Claude to assess clinical-trial plans for an early-stage drug and estimate their likelihood of success. Michael Pourdehnad, MD, the company's senior vice president and head of early clinical development for hematology, oncology and cell therapy, is quoted: "From a single prompt, Claude generated a robust initial analysis." Anthropic said the collaboration is planned to expand in 2027.

The relationship predates the post. Reuters reported on May 20 that BMS would give Claude to more than 30,000 employees. Greg Meyers, BMS's chief digital and technology officer, said: "Most enterprise AI stops at the chatbot."

Mithril's founder, Cody Tranbarger, used Claude to coordinate contracts and experiments with 11 external labs, and a possible treatment went from idea to promising mouse results in under 18 months, Anthropic said. Tranbarger now works at Anthropic. If candidates hold up, a drug could be ready for human trials as early as late 2028. Anthropic's separate legal exposure is covered in the Cruz v. Anthropic story, and a rival lab's agent safeguards in OpenAI's training-lockdown report.