Anthropic is shifting Claude from computational biology into physical experimentation, with over 5,000 protein designs now undergoing wet-lab validation and new tools enabling AI agents to operate laboratory equipment. The company announced a coordinated push across four major initiatives in mid-September 2026, marking a transition from predicting biological outcomes to actually testing them in real laboratories.

What Is Anthropic Actually Building in Biology?

Anthropic's biology strategy centers on closing the gap between what AI can predict and what scientists can verify experimentally. In just under four weeks, Claude optimized more than 30 open-source deep learning models used for biological tasks, speeding them up roughly 4 times on average while maintaining accuracy. These models handle structure prediction, protein design, genomics, and protein language analysis. Claude also created a specialized "Big" mode that can predict biomolecular systems larger than 10,000 tokens on a single NVIDIA GPU node, and systems exceeding 70,000 tokens using one NVIDIA B300 node.

The real breakthrough is the validation loop. Anthropic is co-sponsoring a protein design competition with Adaptyv Bio, committing up to $1 million in Claude credits and $250,000 in Modal compute credits, plus wet-lab validation for over 5,000 designs. DNA for these designs is being provided by Twist Bioscience. This means Claude's predictions are not staying in silico; they are being synthesized and tested in actual laboratories.

How Is Anthropic Enabling AI to Control Lab Equipment?

On August 27, 2026, Anthropic introduced the Model Hardware Standard (MHS), a specification that allows AI agents to safely discover, read from, and operate physical laboratory and manufacturing equipment. The standard is model-agnostic, meaning it works with any device that has a programmable interface. This includes microscopes, robotic arms, and liquid handlers, enabling tasks ranging from routine drug discovery experiments to laser calibration on quantum computers.

The MHS opened as a research preview to an initial group of labs and manufacturers, signaling that Anthropic is moving beyond theoretical collaboration into hands-on partnership with research institutions and industry partners.

What New Tools and Programs Did Anthropic Launch?

• Claude Science Workbench: Introduced June 30, 2026, this customizable app integrates tools and packages researchers commonly use, produces auditable artifacts, and provides flexible access to computing resources for scientific workflows.
• Life Sciences Verification Program (LSVP): Launched in beta on September 17, 2026, LSVP gives life science professionals access to Anthropic's Mythos, Opus, and Sonnet models with refined safeguards tailored for biology-related work. Applicants undergo verification of research credentials, security standards, and ethical research oversight. The program enables work currently blocked in general-availability models, including drug discovery, research biology, clinical development, and manufacturing.
• Expanded AI for Science Support: Researchers can now apply for up to $50,000 in credits per project, with 10,000 one-year Claude Team plan seats available for academic and nonprofit labs.

Who Is Already Using Claude for Drug Discovery?

On September 16, 2026, pharmaceutical company Novo Nordisk and Anthropic announced a collaboration to accelerate drug discovery. Novo is testing Claude and Claude Science in selected research and development workflows, and using Anthropic's frontier models to strengthen AI-driven software development. This partnership demonstrates how a major enterprise is integrating Claude into real biological research pipelines, not just pilot programs.

The Novo Nordisk collaboration is significant because it shows a governed, production-level use case. Novo's scientists and computational teams have identified specific biological reasoning workflows where Claude is being tested, suggesting the company sees practical value in the models for its core business.

Why Does Safety Matter in This Expansion?

Anthropic has paired its expanded capabilities with explicit safeguards. The Life Sciences Verification Program requires applicants to demonstrate research credentials and ethical oversight, preventing misuse while enabling legitimate biological research. In September 2026, Anthropic published a report titled "Detecting and countering misuse of AI," which details misuse cases including biological misuse, underscoring why verification and safeguards are paired with expanded capability.

The verification approach reflects a deliberate strategy: give capable tools to credentialed researchers under oversight, rather than blocking biology work entirely. This balances innovation with responsibility, allowing drug discovery and clinical development while maintaining guardrails against harmful applications.

What Does This Mean for the Future of AI in Science?

Anthropic's announcements reveal an emerging experimental loop. Computational prediction and design are optimized by Claude, validated through external wet-lab programs, designed to let AI agents operate laboratory hardware, and gated through a verification program for sensitive biology work. This is not a single laboratory or isolated capability; it is a coordinated ecosystem spanning model optimization, hardware control, wet-lab validation, and enterprise partnerships.

The significance lies in translating computational predictions into experimentally validated results. Over 5,000 protein designs moving from Claude's predictions into physical synthesis and testing represents a tangible shift from theory to practice. If these designs validate at scale, it could demonstrate that AI-driven biology workflows can produce real scientific outcomes, not just promising simulations.