Anthropic is entering drug development directly, targeting neglected diseases the pharmaceutical industry considers unprofitable. Alongside the launch of Claude Science, a dedicated AI workbench for researchers, Anthropic announced early-stage drug discovery programs — while Novartis CEO Vas Narasimhan says AI could cut development timelines from 12 years to 7-8 and double success rates from 8% to 16%.
The Decoder
Anthropic launches its own drug discovery programs to tackle diseases Big Pharma considers unprofitable
STAT
AI company Anthropic announces it will begin developing drugs of its own
TechCrunch
Anthropic's Claude Science bets on workflow, not a new model, to win over scientists
anthropic.com
Claude Science, an AI workbench for scientists
Anthropic has announced it will launch its own drug development programs, focusing on diseases the traditional pharmaceutical and biotech industries consider unprofitable. The move was revealed at the company's AI for Science event, which also served as the launch platform for Claude Science — a dedicated AI workbench for researchers. Anthropic says the drug programs align with its nonprofit mission and will help it build better AI tools through hands-on scientific experience. The company plans to operate at the preclinical stage, stopping short of commercialization — at least for now.
Claude Science is not a new AI model but a purpose-built research environment: it gives scientists one place to run computational work across more than 60 scientific databases, manage compute on local or remote infrastructure, and produce fully auditable, reproducible artifacts. Early users include researchers at the Allen Institute and UCSF's Brain Tumor Center, who report dramatic time savings in tasks like germline analysis and long-form literature reviews.
Novartis CEO Vas Narasimhan offered a concrete breakdown of AI's potential impact at the same event. Currently, getting a drug candidate from development to approval takes about 12 years. Narasimhan split delays into three buckets: information latency, operational latency, and biological latency. AI tools could sharply reduce the first two — which account for roughly 40% of total development time — bringing timelines down to 7 or 8 years.
On success rates, Narasimhan sees room to double the current 8% approval rate to 16%, driven by better safety predictions and optimized molecular properties. The challenge that remains hardest: confirming whether a drug target is biologically correct for a given disease. Even so, he argues the scale impact is enormous — major pharmaceutical companies collectively spend $150–$200 billion per year on R&D and have produced only 800–1,000 approved drugs over 120 years. More diseases could become treatable, and previously unreachable drug targets could become viable.
Anthropic isn't alone in pushing AI into medicine. Google DeepMind owns foundational science models like AlphaFold and AlphaGenome outright — tools no competitor can replicate, only call into as APIs. Its Gemini for Science platform bundles those models with over 30 life science databases. Notably, AlphaFold co-developer John Jumper recently left DeepMind for Anthropic. OpenAI took a different route in April 2026, releasing GPT-Rosalind — a specialized model fine-tuned for biological reasoning, gated behind an enterprise qualification process with partners like Amgen, Moderna, and Novo Nordisk.
The three strategies represent distinct bets: Anthropic is going broad with wide subscription access to Claude Science; OpenAI is going narrow and enterprise-gated with a specialized model; and Google is leaning on proprietary foundational models nobody else has. Google DeepMind also launched an AI Co-Clinician in 2026 for clinical settings, while OpenAI's ChatGPT Health lets users connect medical records and wellness apps. Independent experts, including Oxford's Catherine Pope, urge caution — especially when AI is applied to diagnosis and direct patient care in messy, real-world healthcare environments.
Claude Science is available in beta on macOS and Linux for Claude Pro, Max, Team, and Enterprise subscribers. Team and Enterprise users need admin enablement. Anthropic is also offering a discounted Team plan for active scientific labs at academic institutions and nonprofits. The company will support up to 50 Claude Science research projects with up to $30,000 in credits each — applications are open through July 15, 2026, with projects running September through December 2026.
Anthropic's drug discovery ambitions signal a meaningful expansion beyond being a model provider. Whether the company's direct involvement in neglected disease research produces viable drug candidates remains to be seen — but the move puts it in direct strategic competition with Isomorphic Labs (Google DeepMind's drug discovery spinout, which raised $2.1 billion) and a growing field of AI-native biotech companies.
Anthropic is launching its own drug development programs for neglected diseases and a new AI research workbench called Claude Science, signaling a major push into pharma and life sciences.
