Anthropic has confirmed it built a wet laboratory dedicated to biology research in the San Francisco Bay Area, a facility that has been operating quietly since spring 2026. The company said this week that the lab, which pairs its AI model Claude with human scientists, has already produced its first result: the identification of an enzyme system it suspects could represent a new gene-editing mechanism.

The project was kept under wraps until this week's announcement. The work centered on searching massive databases of DNA sequences for proteins and enzymes that scientists had not yet studied closely.

Anthropic CEO Dario Amodei marked the announcement with a post on X:

"Today we announced the Claude-led discovery of a molecular machine that we suspect could represent a new gene editing mechanism. Its precise function, biotechnological utility (if any), or level of significance is not yet clear, but at minimum it is work I would have been proud… "

— Dario Amodei (@DarioAmodei) via X, September 23, 2026 (post)

How the Discovery Was Made

The search started with a simple prompt asking Claude to look for new reverse transcriptase functions. Reverse transcriptases are enzymes that write RNA into DNA. Anthropic deployed 950 AI agents to run the search over the of a single day, and together they gathered more than 200,000 candidate enzymes.

That list was narrowed down to the 20 most promising candidates. One of them had already been studied before, but the AI agent noticed something new about it: the enzyme contained a repeating pattern of DNA code. Patterns of this kind are linked to how CRISPR systems cut and paste genetic material. Gene-editing technologies such as CRISPR were themselves built on enzymes that bacteria use naturally, a precedent that helps explain why uncharacterized enzyme systems like this one draw close scrutiny from researchers.

Anthropic named the new enzyme system array-associated reverse transcriptases, or ART. The company published a technical report on the finding, though it has not yet been peer-reviewed.

Humans Still Run the Bench Work

Eric Kauderer-Abrams, Anthropic's head of life sciences, spoke with Reuters about the project. He said the facility still looks like a typical molecular biology lab, and that the main difference lies in what happens before physical testing begins. Claude does the early work of scanning sequences and narrowing down candidates, while human scientists then test the top candidates in the lab. Final testing, he noted, still has to happen in real lab conditions rather than on a computer.

According to the company, the ART discovery involved six named human scientists working with Claude, and the project used 210 million tokens to complete the search and analysis.

Outside Researchers Weigh In

Feng Zhang, a CRISPR pioneer who works at MIT and the Broad Institute, commented on the discovery. He described the finding as an example of how AI agents can help with biology research, and added that the link between the enzyme and RNA repeat arrays is worth studying further.

The full function of ART remains unknown. Any confirmation of the suspected gene-editing role — and of any practical use — would rest on the same real-world bench testing Kauderer-Abrams described, along with the peer review the technical report has yet to undergo.

Separation From Drug Development

Anthropic said its research program is separate from drug development, and the company stated that it will not run clinical trials tied to this work. Kauderer-Abrams told Reuters that Anthropic does not see itself as competing with pharmaceutical and biotech companies, noting that those firms focus on bringing drugs to market — which is not Anthropic's goal.

The company does count major pharmaceutical firms among its partners, including Roche, Genentech, and Bristol Myers Squibb. Those companies may use Anthropic's AI tools, although Anthropic's own lab has resources that outside academics do not have access to.

Some researchers have raised concerns about funding in this emerging field. The Scientist reported in July that academics worry about having to choose between paying for AI tokens and supporting trainees.

Wider AI Biology Race

Other companies are also working on AI-driven biology research. Alphabet's Isomorphic Labs — a company Alphabet built out of its DeepMind AI research unit to apply machine learning to drug discovery — and OpenAI have both published work in this space in recent months.

Source: CoinCentral