AI-Driven Discovery
Anthropic confirmed it is operating a wet biology lab in the Bay Area where it uses AI models to support physical experiments. The company announced that this lab has identified what it believes is a major discovery: a new enzyme system with properties reminiscent of CRISPR.
CRISPR is a natural immune system used by bacteria to defend against viruses that has evolved into a widely utilized gene-editing technology for researchers.
The finding consists of a previously unknown enzyme system hidden in the DNA of bacteriophages, viruses that infect and replicate within bacteria. Researchers state that this system behaves similarly to CRISPR by performing operations such as cutting, copying, and pasting DNA.
Community Validation and Context
Independent validation from the broader research community will be required to determine the significance of the finding. CEO Dario Amodei noted that the work built on existing research and that a Stanford team previously discovered a somewhat similar system.
Amodei emphasized that the discovery was generated mostly, though not entirely, by the Claude model.
The lab was established earlier this spring, and the analysis required approximately 21 hours of concerted effort. The model processed data using around 950 agents consuming 210 million tokens.
Safety and Risk Considerations
The revelation follows public statements from AI industry leaders regarding the potential risks of advanced models, leading to calls for slower deployment and formalized safety testing procedures.
Amodei has previously stated concerns about potential bioterrorism risks associated with advanced AI, while also expressing optimism that technology could help cure major diseases within the coming decade.
Human-Led Laboratory Operations
Despite the heavy involvement of AI in data analysis, physical experiments inside the facility are currently conducted entirely by human scientists.
The Bay Area facility operates at lower biosafety risk levels of BSL-1 and BSL-2 and does not handle pathogens capable of infecting humans, with all practical laboratory work performed by human researchers.
Industry Landscape
AI-powered biological research is actively pursued across the technology sector. Stanford researchers recently published work utilizing large language models for CRISPR research, UC San Francisco researchers used AI to design de novo enzymes, and Google previously introduced AlphaFold.
While automated laboratories controlled directly by AI remain a possibility for the future, Anthropic maintains that its current operations rely on strict human supervision and safeguards.
.
