Anthropic launches Claude Science: AI for drug discovery and labs

Anthropic announced Claude Science at an event for pharmaceutical executives, biotech founders, and researchers yesterday, unveiling a flagship product designed to support scientific research. The company says Claude Science can carry out meaningful work from concise, high-level instructions and includes tools aimed at computational biology and drug development; Anthropic will also use it in its own research into drugs for rare, neglected diseases.

What is Claude Science and how does it work?

Claude Science is a new, science-focused version of Anthropic’s assistant technology that mirrors how Claude Code supports software engineering: it takes concise, high-level instructions and performs autonomous, substantive tasks. The product ships with tools specifically for computational biology and drug development, positioning it as an assistant for laboratory planning, data analysis, or other workflows where researchers need domain-specific automation.

Beyond the headline features, Anthropic framed Claude Science as both a commercial product and an internal research tool. The company intends to apply the platform to its own drug-discovery work for rare and neglected diseases, signaling a dual use case: selling the product to biotech and pharma teams, and using it to accelerate internal projects.

How does California’s dairy methane program work, and what’s the problem?

California’s Low Carbon Fuel Standard lets petroleum companies buy credits from cattle farmers that capture methane from manure, and convert that biogas into pipeline-quality natural gas; those credits substitute for direct reductions in fuel carbon intensity. The state assumes methane has about 25 times the warming effect of carbon dioxide over a 100-year horizon, and regulators count captured biogas as avoiding powerful short-term methane warming while permitting more CO2 from burning the gas.

That accounting produces large incentives. UC Berkeley economist Aaron Smith calculated that adding one average biogas-powered vehicle to a fleet would generate enough LCFS credits to cover deficits from 26 similar gasoline-powered vehicles. Critics say the math mixes short-lived and long-lived greenhouse gases in ways that reduce near-term warming at the cost of adding carbon dioxide that persists for centuries. California regulators extended parts of the program beyond 2050 in 2024, and a recent proposal by the state’s air resources board could channel millions of additional dollars to dairy farmers while easing restrictions on major greenhouse-gas producers.

Why it matters

Anthropic’s push into scientific workflows shows how vendors are packaging large models not just to write code or copyedit text but to perform domain-specific research tasks; the company’s plan to use Claude Science in its own rare-disease drug work highlights commercial interest and internal dependence on these systems. At the same time, the dairy-methane case underscores how policy choices about measurement and incentives can lock in long-term climate outcomes. Both stories expose a common theme: complex technical tools—whether AI systems or carbon accounting mechanisms—are being deployed rapidly, and the downstream effects depend on how accurately designers and regulators map short-term gains to long-term consequences.

What to watch

Watch whether Anthropic publishes examples or benchmarks showing Claude Science performing end-to-end drug-discovery tasks, and whether the company discloses outcomes from its own rare-disease projects. On the policy side, monitor the state air resources board’s proposal and any final decisions about funding or relaxed restrictions, since the plan could send millions more to dairy farms and shape how LCFS credits continue to be used.