Anthropic Launches Claude Science, Enters Drug Discovery With Its Own Pipeline for Neglected Diseases

On June 30, 2026, Anthropic did something that surprised even close observers of the AI industry: it announced not just a new research product, but an internal drug development program—signaling that the company founded to make AI safe now believes that AI-native pharmaceutical development is within its reach and mission.

The dual announcement, unveiled at “The Briefing: AI for Science” event in San Francisco, bundled Claude Science—a unified research workbench for scientists and pharma companies—with a separate but related commitment to develop Anthropic’s own drug candidates targeting neglected tropical diseases. It is a significant pivot for a company that has largely stayed out of vertical product markets, and it positions Anthropic squarely in a three-way race against OpenAI, which has partnered with Novo Nordisk, and Google DeepMind’s Isomorphic Labs, which raised $2 billion for AI-driven drug discovery.

What Claude Science Actually Is

Claude Science is built atop Anthropic’s existing Claude models—primarily Sonnet 5 and Fable 5—but connects them to a purpose-built research environment designed to solve a problem any bench scientist can describe: the constant context-switching between dozens of specialized databases, file formats, and analysis tools that currently consumes a disproportionate fraction of research time.

The platform integrates more than 60 scientific databases spanning genomics, proteomics, structural biology, cheminformatics, and single-cell analysis. Researchers can work in a single interface to query UniProt for protein structures, run cheminformatics analysis on molecular datasets, cross-reference clinical trial registries, and interrogate genomic databases—without needing to manually export, reformat, and reimport data between systems.

According to MIT Technology Review, which was briefed on the product before launch, the core value proposition is context persistence: Claude Science maintains awareness across a multi-hour research session, allowing a researcher to ask increasingly specific follow-up questions about a molecule or disease pathway without restating the full experimental context each time. The company describes it as providing “the kind of intellectual continuity a very senior scientific collaborator would provide.”

Claude Science is currently in beta for Pro, Max, Team, and Enterprise plan subscribers. Anthropic is also offering research credits worth up to $30,000 for projects exploring the boundaries of AI-driven science, with applications open through July 15.

The Internal Drug Program: Deliberate and Unconventional

More surprising than the product is what accompanied it. Anthropic announced it is establishing an internal drug discovery program—distinct from the Claude Science product—that will identify and develop its own drug candidates targeting neglected tropical diseases: conditions like leishmaniasis, Chagas disease, and African sleeping sickness that afflict hundreds of millions of people in low-income countries but generate insufficient commercial returns to attract major pharmaceutical investment.

The rationale, CEO Dario Amodei explained in remarks at the San Francisco event, is that these diseases represent a test case where frontier AI could genuinely accelerate the discovery timeline without the commercial pressures that shape mainstream pharma R&D. Neglected diseases have well-characterized biology, modest regulatory complexity by global standards, and represent genuine unmet need—making them an appropriate domain for an AI company to explore without competing directly with established industry players.

Anthropic’s internal program will use Claude’s models to identify candidate compounds, analyze structural biology, model protein-ligand interactions, and generate hypotheses about drug mechanisms. The company has not yet disclosed a target number of drug candidates or a timeline to clinical trials, but it has committed to publishing research findings regardless of commercial outcome—a notable stance in an industry where negative results are historically suppressed.

The Three-Way Race in AI Drug Discovery

The announcement drops Anthropic into a competitive landscape already defined by two heavyweight incumbents.

Google DeepMind’s Isomorphic Labs, spun out of DeepMind’s AlphaFold team in 2021, closed a $2 billion funding round in mid-2026 and is in active collaborations with Eli Lilly and Novartis. Its AlphaFold 3 model can predict the structure of protein-ligand complexes with near-experimental accuracy, giving it a significant head start in structure-based drug design.

OpenAI, meanwhile, has embedded AI into drug discovery through its partnership with Novo Nordisk and its investment relationship with Elucidate Health and several smaller biotech startups. OpenAI’s approach has focused on applying general-purpose o3-class reasoning models to drug design problems, rather than training bespoke scientific models.

Anthropic’s positioning differs on both dimensions: Claude Science is a general-purpose scientific workbench rather than a drug-discovery-specific tool, while the internal program targets diseases that neither Google nor OpenAI has publicly committed to pursuing. Whether that combination—broad scientific platform plus narrowly targeted proprietary program—is a coherent strategy or a stretch remains to be tested.

Implications for Pharma Partnerships

The entry of Anthropic into life sciences also has implications for how major pharmaceutical companies will structure their AI partnerships over the next two years. The three frontier AI labs now all have credible scientific products, which should accelerate institutional adoption but also raise questions about data sharing: will pharma companies be comfortable running their most proprietary compound libraries through a platform operated by a company that also has its own drug pipeline?

Anthropic has acknowledged this tension and says Claude Science operates under strict data isolation guarantees, with enterprise customers’ data never used to train models or shared across organizations. Whether that commitment is sufficient to unlock the most sensitive research workflows—where competitive advantage hinges on preclinical data that can be worth hundreds of millions of dollars—will be one of the key determinants of Claude Science’s commercial success.

For now, the announcement establishes something that would have seemed implausible three years ago: the three companies most central to the frontier AI race are also all, in different ways, in the business of developing medicine. The speed at which AI has moved from abstract capability to biological research tool is itself a data point about the pace of this technology’s diffusion into the fabric of every knowledge industry.