Anthropic AI Identifies New Enzyme System in DNA

On September 24, 2026, Anthropic, a U.S. artificial intelligence startup, announced a significant breakthrough from its molecular biology lab. The company's Claude AI agents identified a previously unknown enzyme system, dubbed ART, within bacteriophage DNA sequences. This finding marks the first discovery made by Anthropic’s AI technology, highlighting its potential to enhance and expedite basic research in molecular biology. Identifying novel enzyme systems is critical for advancing our understanding of biological processes and could have far-reaching implications in biotechnology and medicine.

The discovery resulted from deploying approximately 950 Claude-powered software agents, which analyzed a comprehensive database of DNA sequences over 21 hours. During this analysis, the agents uncovered over 200,000 genes associated with one type of enzyme, ultimately narrowing their focus to 20 candidates worthy of further investigation. This process, which would typically require weeks or even months of work by human experts, was completed in a remarkably short timeframe, underscoring the efficiency of AI in research settings.

Anthropic's researchers provided initial instructions and conducted lab experiments, while the AI agents autonomously determined which leads to pursue. One notable finding was a set of short, repeating DNA sequences adjacent to two genes: one coding for a known enzyme and the other for an unidentified protein. This configuration resembles the CRISPR system, a natural defense mechanism in bacteria that has been adapted for gene editing in medical applications. The potential for ART to serve as a new gene-editing mechanism is being closely examined, with implications for gene therapy in the future.

The preliminary nature of this discovery was highlighted by Anthropic's chief, Dario Amodei, during his address to the UN Security Council. He emphasized that while the findings are in the early stages, they may pave the way for significant advancements in treating diseases that have long challenged humanity. Initial experiments suggest that the ART system could be programmable, similar to CRISPR, raising exciting possibilities for its application in genetic engineering.

Feng Zhang, a leading figure in CRISPR research, described the identification of the repeated DNA fragments as genuinely intriguing, indicating that it merits further investigation. Additionally, Amodei noted that a team at Stanford University has independently described a comparable, yet distinct, system. Anthropic's lab, located in the San Francisco Bay Area, adheres to strict safety protocols, ensuring that experiments do not involve pathogens that could infect humans. The potential for Claude to operate lab equipment autonomously in the future is also being considered, contingent on implementing proper safeguards.