Anthropic pairs Claude with six drug developers and starts its own discovery programs

Anthropic has spent the past several months working directly with six organizations that develop medicines — BigHat Biosciences, Bristol Myers Squibb, Divergence Labs, Enveda, Generate Biomedicines and The Column Group — and has started drug discovery efforts of its own, the company said in a research post published Tuesday. It plans to expand the effort significantly from here.

The announcement, framed as an early progress report rather than a product launch, gives the clearest picture yet of how Anthropic intends to apply Claude in an industry defined by long timelines and heavy documentation.

A generalist strategy rather than a single chemistry tool

According to Anthropic, drug development contains many separate bottlenecks — choosing targets, designing molecules, making them manufacturable at scale, drafting regulatory filings, designing and running trials — and there is no single fix. Its approach is therefore to embed Claude across the whole chain rather than optimize one task.

Fathom News, covering the announcement, argued this distinguishes Anthropic from vendors pitching narrow gains such as faster literature review or early molecule ranking: the bet is that stringing together many medium-value automations, especially paperwork and coordination, is what actually moves real programs.

Early results across the pipeline

Anthropic's post describes results at several stages of development.

Generate Biomedicines is trying to extend CAR T-cell therapy, which has historically worked against blood cancers, to solid tumors. Six of its scientists used Claude Code to redesign part of the therapy's genetic instructions to make room for additional anti-cancer instructions. One proposed design was roughly 70% shorter than the standard sequence; in lab tests, immune cells carrying it produced about ten times as many cancer-detecting receptors, killed tumor cells more efficiently, and remained active longer.

Divergence Labs is pursuing a food allergy treatment that would strip IgE molecules off immune cells rather than only neutralize those circulating in blood. Claude converted the company's goals into a detailed design plan and then, mostly without human intervention, produced around 35,000 antibody designs in about a month. Early screening yielded three designs that bind IgE and twenty that dislodge it, three of which appeared better at dislodging it than the standard medication.

BigHat Biosciences, which pairs AI with an automated lab, worked with Claude to optimize two antibody-fragment leads. About 1,010 variants were made and tested in roughly ten weeks. Midway through, the team suspected the standard heat-stability and clumping metrics were misleading; after Claude helped reanalyze the raw data and propose new metrics, BigHat's scientists ran a revised protocol and the predictions held. The best variants bound more than ten times more tightly than the originals, one lead stayed in solution at temperatures 18°C higher, and the other clumped over 100 times more slowly.

Enveda focused on manufacturing and regulatory paperwork, where chemists had been spending up to 75% of their time on documentation. Across 1,061 pages of manufacturing records, Claude caught all 11 issues human reviewers had identified and seven they had missed, cutting review time by 89%, from over 25 hours to under three. Claude also drafted two regulatory documents: one took four hours instead of roughly three months by hand, and the other surfaced one minor error under Enveda's normal review.

The Column Group is taking a different route, setting up a new company, Pleiades, that will use Claude to run more — and more complex — drug programs than a small team typically could, potentially a template for reducing both scientific and administrative overhead across the industry.

Anthropic's own programs and the caveats

Anthropic has also begun internal drug discovery efforts, including programs for rare childhood forms of muscular dystrophy in which Claude does most of the scientific and operational work. The company says these serve three purposes: exposing model weaknesses, producing case studies for the wider life sciences community, and, if positive trajectories continue, yielding medicines for conditions including food allergies, rare childhood diseases and several forms of cancer.

Fathom News noted that Anthropic disclosed no commercial terms, equity stakes or revenue tied to the partnerships, and did not say whether any Claude-assisted work has entered formal clinical testing. Claude is also not operating lab equipment; its role is planning, drafting, comparison and decision support between scientists, operators and regulators.

Why it matters

The announcement marks a shift from benchmark claims to applied AI inside regulated, high-stakes workflows. Anthropic is positioning Claude as a general worker across entire drug programs rather than a better molecule generator, wagering that compounding gains across many tasks — not isolated chemistry wins — will compress timelines. The thesis carries risk: as Fathom News points out, small errors at one stage can compound months later in medicine, and the near-term test will be concrete outputs such as molecules advancing into preclinical work and partner-verified cycle-time reductions. For now, the published results are early but unusually specific for an AI vendor entering biotech.