Anthropic has announced a groundbreaking discovery made by its AI system Claude while analyzing genomic data of bacteriophages — viruses that infect bacteria. As part of the Life Sciences program, a network of approximately 950 AI agents processed over 200,000 fragments of biocode in 21 hours and identified a previously unknown biological structure named ART (array-associated reverse transcriptase). This was reported by RBC-Ukraine on September 24, 2026, citing a post by Anthropic CEO Dario Amodei on the social network X.

What the ART System Is

According to the description provided in Anthropic's materials, the ART structure combines three key components: a specific enzyme — reverse transcriptase, a neighboring partner gene, and a long chain of identical DNA repeats located at equal distances from one another. It is precisely this architecture — the alternation of repetitive sequences near the enzyme gene — that makes ART structurally related to known CRISPR systems, which bacteria use to recognize and neutralize foreign DNA, and which scientists employ for genome editing. At the same time, the exact biological function and mechanism of action of ART in the life cycle of bacteriophages remain unknown at the time of publication, which, according to the researchers, opens up a direction for further experimental investigation.

Methodology: How the AI Conducted "Genomic Mining"

To carry out the large-scale analysis, Anthropic deployed a network of around 950 AI agents that performed so-called "genomic mining" — the automated search for anomalous patterns in vast arrays of genetic sequences. The algorithms processed more than 200,000 reverse transcriptases, from which they isolated about 3,500 potential systems with characteristics typical of defense mechanisms, and ultimately narrowed the field down to 20 most promising candidates for the preparation of detailed scientific reports. The total volume of computations amounted to approximately 210 million tokens. According to Anthropic's own estimates, performing a similar volume of analytical work by a professional scientist would have taken several weeks or even months of painstaking effort.

The Key to the Discovery: Attention to Anomaly

As follows from the description of the process, the turning point came when one of the AI agents noticed an atypical group of genes and conducted an in-depth analysis of the DNA region in its immediate vicinity. The algorithm detected an unusual set of repetitive fragments located next to the reverse transcriptase gene. This pattern had previously gone unnoticed in open genomic databases, which, in the developers' view, demonstrates the ability of AI to identify unknown biological mechanisms in data whose volume exceeds the capacity of manual analysis.

Context: The Life Sciences Program and the Role of AI in Biology

The discovery was made within the framework of Anthropic's new biological program for analyzing DNA sequences, which the company positions as a tool for accelerating fundamental research. The involvement of hundreds of autonomous AI agents in tasks traditionally performed by molecular biology laboratories underscores the growing role of artificial intelligence in genomics. Previous examples of the application of CRISPR-like bacteriophage systems — in particular, the 2021 experiments in which a modified phage carrying a CRISPR module selectively eliminated a specific strain of E. coli in mice — show that understanding new viral defense mechanisms may have direct biotechnological applications. However, in the case of ART, practical utility has not yet been established: researchers emphasize that it is first necessary to determine exactly what function this structure performs in the life cycle of the bacteriophage.

Contradictory Data

No direct contradictions between sources on the key figures (21 hours, 200,000 transcriptases, 950 agents, 210 million tokens) have been identified. At the same time, it should be noted that information about the discovery is currently based on Dario Amodei's post on X and a retelling by RBC-Ukraine; independent peer review in scientific journals has not been confirmed at the time of publication of this article. Moreover, the original RBC-Ukraine text contains a passage in which the formulation regarding the significance of the discovery and its biotechnological usefulness is vague and does not include specific data on planned experiments. Thus, the fact of detecting the ART pattern in genomic data is confirmed by the primary source, however, its biological interpretation and practical significance remain subject to further scientific validation.