Anthropic Launches Claude Science and Enters Drug Discovery

Anthropic has moved directly into scientific research, launching a new flagship product called Claude Science and announcing that it will use the tool to run its own drug discovery program.

The company is doing something unusual for an AI lab: instead of just selling tools to researchers, it plans to use them to hunt for drugs itself — starting with diseases the pharmaceutical industry has largely ignored.

What Claude Science actually is

Claude Science is a research workbench — a single environment that bundles the software tools and data connections scientists reach for most, so they don't have to stitch them together by hand.

It integrates more than 60 preconfigured tools and connectors into one place, produces auditable artifacts (records of exactly what the AI did, so results can be checked and reproduced), and provides flexible access to computing power — local, remote, and high-performance.

Why this matters: real scientific work is a mess of disconnected software, custom scripts, and heavy computation. Pulling it into one auditable environment is the difference between an AI that can chat about science and one that can actually help run it.

The product is available now to all paid Claude subscribers, putting it in reach of individual researchers, not just large institutions.

From tool vendor to drug hunter

The bigger surprise is what Anthropic plans to do with it. The company says it intends to use Claude Science to pursue its own internal drug discovery program.

The focus is deliberate: neglected diseases that traditional pharmaceutical companies have historically found commercially unattractive. These are illnesses — often concentrated in poorer parts of the world — where the potential profit doesn't justify the enormous cost of conventional drug development.

Why this matters: neglected diseases are a textbook market failure. The science may be tractable, but the economics don't work for big pharma. If AI can slash the cost of early research, it changes which diseases are worth pursuing in the first place.

Why an AI company would do this

On the surface, an AI lab running a drug program looks like a strange detour. But there's a logic to it.

Drug discovery is one of the hardest, highest-value tests of whether an AI can do real scientific reasoning — not recite biology facts, but analyze data, form hypotheses, and guide experiments. Succeeding there is a powerful proof of capability.

Choosing neglected diseases sidesteps the awkwardness of competing head-on with the pharma companies Anthropic also wants as customers. It's a mission-friendly wedge into a field where the commercial incumbents aren't really trying.

Backing outside researchers, too

Anthropic isn't keeping Claude Science to itself. The company is supporting up to 50 outside "AI for Science" projects, offering up to $30,000 in Claude credits each. The cloud-compute firm Modal is chipping in up to $2,000 in compute for select projects.

That's a modest sum by frontier-AI standards, but it's aimed at exactly the people who could stress-test whether Claude Science works: academic and independent researchers with real problems and limited budgets.

Who this affects

For working scientists, Claude Science is a new option for doing computational research without assembling a toolchain from scratch — and it's available on a normal paid subscription.

For biotech and pharma, Anthropic's move is a signal. A leading AI lab now believes its models are good enough to attempt drug discovery directly. Whether that's a threat or a future partner depends on how the internal program performs.

For patients affected by neglected diseases, it's a rare bit of attention. If AI genuinely lowers the cost of early-stage research, conditions long deemed unprofitable could finally get a serious look.

What's next

The near-term question is credibility. Announcing a drug discovery program is easy; producing a validated candidate is not. Watch for concrete research output — published findings, identified compounds, or partnerships with wet-lab organizations that can test what the AI proposes.

Also watch adoption. If independent researchers start doing real work in Claude Science and sharing reproducible results, that's the strongest evidence the workbench is more than a well-packaged demo.

The bottom line

Anthropic just did two things at once: shipped a research workbench for scientists and declared it will use that workbench to discover drugs itself.

If you're a researcher, Claude Science is worth trying — it's available on paid plans today. If you're watching the AI industry, the drug program is the story to track. It's a bet that AI is ready for real science, and the results, not the announcement, will tell us whether that bet pays off.