Anthropic Runs Biology Lab for AI Experiments

Anthropic is operating a biology laboratory that conducts experiments, according to TechCrunch, a notable move for an AI company best known for building large language models and emphasizing safety. The lab comes at a moment of contradiction in the AI industry: leading figures promise that AI will help cure diseases and transform medicine, while Anthropic’s own researchers have warned that advanced AI could pose existential risks. The biology lab sits directly in that tension. It likely serves multiple purposes. One is to explore how AI can accelerate biological research, from analyzing experimental data to designing proteins, compounds, or research workflows. Another is to study the safety implications of applying AI to biology, including dual-use risks, biosecurity concerns, and the possibility that automated systems could lower barriers to harmful experimentation. By running its own lab, Anthropic can gain firsthand experience with the realities of biological work rather than relying only on theoretical models. That could improve how it evaluates AI systems for biological capabilities and risks. It may also help the company build better safeguards, such as monitoring for dangerous queries, controlling access to sensitive tools, and testing whether models can provide meaningful assistance to experts. The move reflects a broader trend of AI labs moving closer to scientific domains. Google DeepMind, for example, has pursued protein structure prediction and materials science. But a physical laboratory is a different kind of commitment, requiring specialized staff, equipment, protocols, and safety oversight. It also raises questions about governance. Will Anthropic publish results, share data, or collaborate with outside biologists? How will it manage biosafety and security? And will commercial incentives conflict with caution? The dual narrative around AI—salvation and danger—makes Anthropic’s biology lab a significant experiment not only in science but in institutional design. If handled transparently, it could become a model for responsible AI-driven biology. If not, it may intensify debates about whether private AI companies should be conducting sensitive biological research at all.