Anthropic recently confirmed that its dedicated biology facility in the San Francisco Bay Area is generating major scientific results. Company leadership announced that the laboratory uncovered an enzyme system with traits similar to CRISPR, opening fresh avenues for genetic manipulation research.

CRISPR functions as a natural defense system that bacteria use to fend off viral invaders. Geneticists widely use the technology to edit, modify, and rewrite DNA sequences with high precision. Anthropic’s team discovered a previously unknown enzyme mechanism hidden deep inside the DNA of bacteriophages, which are specialized viruses that infect bacterial cells. This newly identified viral system performs fundamental genetic tasks, including cutting, copying, and pasting target DNA strands.

The broader scientific community must validate these findings before ranking their overall impact. Anthropic CEO Dario Amodei noted that this discovery builds on earlier research done by outside teams, including a Stanford group that previously identified a related genetic system.

Anthropic highlights the incredible speed of the discovery process. Although the Bay Area biology lab opened earlier this spring, its computational model did the heavy lifting in just 21 hours of focused processing time. The Claude model searched through massive biological datasets using approximately 950 software agents, consuming roughly 210 million tokens during the analysis run.

Running physical biology experiments alongside computational models sparks intense industry debate. Amodei previously warned that advanced models could pose biosecurity risks if bad actors misuse automated research tools. Other researchers inside the company openly stated that uncontrolled software models could introduce severe safety hazards. Amodei himself acknowledged risks around bioterrorism, but emphasized his belief that intelligent research systems will cure most human diseases over the next decade.

To manage safety concerns, Anthropic enforces strict protocols inside its physical laboratory. The company clarified that Claude does not run physical experiments autonomously. Human scientists conduct all physical laboratory work inside biosafety level one and two facilities, strictly avoiding dangerous human pathogens.

Using software models to accelerate biological discovery is gaining traction across top research hubs. Stanford scientists recently published papers detailing molecular biology models, while researchers at the University of California, San Francisco used software models to design brand-new enzymes from scratch. Meanwhile, Google’s AlphaFold continues serving as a core tool for global protein structure research.

While human staff currently run all lab equipment, Amodei hasn’t ruled out automated physical testing down the line. As safety protocols mature, future setups may allow models to execute physical experiments directly under strict safety controls. For now, human hands handle the physical test tubes, while intelligent models hunt through mountains of genetic code to uncover nature’s hidden mechanisms.