Artificial intelligence company Anthropic is moving from computer screens into the physical laboratory.
The Claude developer has quietly established a biology wet lab in the San Francisco Bay Area, giving its researchers the ability to conduct physical biological experiments as the company expands its ambitions in medicine, biotechnology and AI-assisted drug discovery.
The move represents an important shift in the AI race. Companies have spent years building models capable of analysing scientific papers, writing computer code and interpreting enormous datasets. Anthropic is now exploring what happens when AI becomes connected more directly to real-world biological experimentation.
Why Anthropic Built a Physical Biology Lab
Much of modern AI-powered biology happens in silico — through computer simulations, biological databases and computational models.
But biology ultimately has to work in the physical world.
A computer model can predict that a particular molecule or protein design should behave in a certain way. Scientists still need laboratory experiments to determine whether that prediction is correct.
Anthropic's head of life sciences, Eric Kauderer-Abrams, told Reuters that real laboratory work remains the final test in biology. The company's new facility allows its researchers to connect AI-generated ideas with physical experiments.
Claude Is Moving Into Drug Discovery
Anthropic's ambitions extend beyond creating a better scientific chatbot.
The company wants AI systems to help researchers investigate treatments for diseases, including rare conditions and biological targets that conventional pharmaceutical development has struggled to address.
Drug discovery is notoriously expensive and time-consuming.
Researchers may investigate thousands of possible compounds before identifying a promising candidate. That candidate must then undergo extensive laboratory testing, preclinical development and eventually human clinical trials before it can become an approved medicine.
AI companies believe advanced models could accelerate parts of that process by helping scientists analyse data, generate hypotheses and prioritise the most promising experiments.
Claude Is Already Working With Scientists
Anthropic has been steadily building specialised tools for scientific research.
Its Claude Science workbench, launched earlier this year, integrates scientific databases, computing tools and research workflows into a single environment designed for scientists. Anthropic says the system can work with tools such as PubMed, Jupyter and scientific computing resources.
The company has also developed Claude models with more advanced capabilities in biology.
On September 17, Anthropic announced a Life Sciences Verification Program, providing approved researchers with access to models including Mythos, Opus and Sonnet under safeguards designed specifically for legitimate advanced biological research.
AI Is Already Accelerating Protein Research
One of the clearest demonstrations of Claude's scientific capabilities involves biomolecular modelling.
Anthropic reported this week that Claude helped optimise more than 30 open-source biomolecular models in under four weeks, making them approximately four times faster on average.
The company also developed a lower-memory approach allowing researchers to model larger biological systems using a single Nvidia GPU node.
That work illustrates why AI companies see biology as such an important frontier.
Scientific research frequently involves enormous datasets, complicated software pipelines and repetitive computational work. AI systems capable of operating those tools can potentially reduce the amount of time scientists spend on routine technical tasks.
Anthropic Is Working With Novo Nordisk
The pharmaceutical industry is also becoming increasingly interested in Claude.
Novo Nordisk announced a partnership with Anthropic this week aimed at using Claude to accelerate drug research and development.
The collaboration demonstrates that AI-driven pharmaceutical research is moving beyond experimental technology demonstrations.
Large pharmaceutical companies increasingly want AI systems that can analyse biomedical information, support researchers and potentially shorten portions of the drug-development process.
The economic incentive is enormous.
Bringing a successful medicine from discovery to market can require years of research and substantial investment. Even modest reductions in development time could therefore have significant financial and medical consequences.
AI Could Connect Directly to Laboratory Robots
The next stage may be even more significant.
Anthropic is exploring how Claude could work with laboratory automation and robotics, allowing AI-generated research decisions to connect more directly with physical experiments. Reuters reported that the company's broader programme includes efforts to automate laboratory processes while retaining human oversight.
Conceptually, such a system could operate as a scientific loop.
An AI analyses existing evidence and proposes an experiment. Automated laboratory equipment performs it. The resulting data returns to the AI, which analyses what happened and proposes the next experiment.
Repeating that cycle could potentially allow researchers to test hypotheses much faster than conventional manual workflows.
This Does Not Mean Claude Is Independently Creating Medicines
There is an important distinction.
Anthropic is not currently conducting human clinical trials, and the existence of an AI-powered biology laboratory does not mean Claude can independently invent an approved medicine.
Drug development remains heavily regulated and requires extensive scientific validation.
Anthropic's current work is concentrated primarily around research and preclinical science rather than replacing pharmaceutical companies or medical regulators.
Human scientists also remain involved in supervising the work.
That matters because biological experimentation can have consequences that ordinary software experimentation does not.
Biology Creates Serious AI Safety Questions
The same capabilities that could help scientists develop treatments can create risks if used irresponsibly.
Anthropic itself acknowledges that advanced biological capabilities are dual use: technology useful for legitimate scientific research may also provide information relevant to harmful biological activity.
That is one reason the company's most capable biology-focused models are not simply available without restrictions.
Its new Life Sciences Verification Program is designed to give vetted researchers greater access while maintaining safeguards around potentially dangerous applications.
The challenge is finding a balance between enabling legitimate researchers and preventing advanced AI capabilities from being misused.
AI Could Change the Economics of Rare-Disease Research
One particularly interesting area is rare disease.
Traditional pharmaceutical economics can make some conditions difficult to prioritise because relatively few patients may need a treatment.
AI could potentially change part of that equation.
If researchers can reduce the cost of analysing biological mechanisms, identifying candidate compounds and designing experiments, diseases that were previously uneconomical to investigate may become more attractive research targets.
Anthropic has specifically identified rare and traditionally difficult-to-drug diseases as areas it hopes AI could help address.
That remains an ambition rather than proof that AI will successfully produce such treatments, but it helps explain why the company is investing in physical laboratory capabilities.
Anthropic Is Becoming More Than a Chatbot Company
The development also illustrates how rapidly the definition of an AI company is changing.
Claude may be best known publicly as an AI assistant, but Anthropic is increasingly building specialised systems for software development, scientific research, healthcare and other professional work.
Its newest Mythos model is specifically positioned for advanced biology and cybersecurity research and is available only to vetted organisations through restricted-access programmes.
Meanwhile, Claude Science provides researchers with an environment designed specifically for scientific workflows.
Adding a physical biology laboratory brings another layer to that strategy.
Anthropic is no longer only building AI that can discuss science. It is building infrastructure intended to help scientists perform science.
The AI Race Is Moving Into the Physical World
For years, the most visible AI competition focused on language models.
Which system could write better? Which could code better? Which could reason across longer documents?
The next stage may be considerably more physical.
AI systems are increasingly being connected to robots, factories, autonomous vehicles and scientific laboratories.
Biology could become one of the most consequential examples because successful research can eventually affect medicines and human health.
But unlike generating text or images, mistakes in biological research can carry real-world consequences.
That makes scientific validation and human oversight essential.
What Happens Next?
Anthropic's wet lab is still part of a developing research programme rather than a pharmaceutical manufacturing operation.
The immediate objective is to connect Claude's computational capabilities with real experimental biology and learn how effectively AI can participate in the scientific process.
Researchers will be watching whether this approach actually produces measurable improvements: faster experiments, better biological models, stronger drug candidates or discoveries that scientists would otherwise have missed.
If it does, the significance could extend well beyond Anthropic.
The biggest competition in artificial intelligence may eventually be measured not only by which model gives the best answer on a screen, but by which AI can help scientists make discoveries in the real world.
