Claude AI Discovers Novel CRISPR-Like Enzyme System in Anthropic’s Wet Lab

Key Highlights

• Anthropic established a biology wet lab staffed by humans in the Bay Area, operating discreetly since spring 2026.
• The facility combined Claude AI agents with human researchers to analyze DNA sequences for undiscovered enzyme capabilities.
• Scientists identified a novel enzyme system named array-associated reverse transcriptases (ART) with CRISPR-like characteristics.
• The breakthrough occurred after deploying 950 AI agents that analyzed more than 200,000 potential enzymes within 24 hours.
• The company has stated it won’t conduct clinical trials and isn’t pursuing competition with pharmaceutical industry players.

Anthropic has verified the establishment of a wet lab focused on biological research. The facility operates in the San Francisco Bay Area and has been functional since this spring.

The initiative remained under wraps until recently. This week, the company revealed that the lab had successfully achieved its inaugural breakthrough.

Claude, Anthropic’s AI platform, was deployed to analyze extensive DNA sequence databases. The objective centered on identifying proteins and enzymes that remained unexplored by the scientific community.

The Path to Discovery

The research began with a straightforward prompt directing Claude to identify novel reverse transcriptase capabilities. These enzymes play a crucial role in transcribing RNA into DNA sequences.

The company deployed 950 AI agents to conduct this analysis within a 24-hour period. Collectively, they identified over 200,000 potential enzyme candidates.

Researchers subsequently filtered this extensive list down to 20 highly promising options. While one candidate had previously received scientific attention, the AI detected a previously overlooked characteristic.

The enzyme displayed a distinctive repeating DNA sequence pattern. Such patterns are characteristically associated with CRISPR mechanisms that edit genetic material.

The company designated this newly identified enzyme system as array-associated reverse transcriptases, abbreviated as ART. Anthropic released a technical document detailing the discovery, although peer review remains pending.

Eric Kauderer-Abrams, who leads Anthropic’s life sciences division, discussed the initiative with Reuters. He noted that the laboratory maintains the appearance of a conventional molecular biology facility.

The primary distinction, he clarified, lies in the preliminary phase before physical experimentation begins. Claude handles the initial sequence scanning and candidate selection process.

Human researchers subsequently conduct laboratory testing on the prioritized candidates. Kauderer-Abrams emphasized that final validation must occur through actual laboratory experimentation rather than computational methods alone.

Future Directions According to Anthropic

Feng Zhang, a CRISPR pioneer affiliated with MIT and the Broad Institute, offered his perspective on the breakthrough. He characterized the discovery as demonstrating AI agents’ potential contribution to biological research.

Zhang noted that the connection between the enzyme and RNA repeat arrays merits additional investigation. The complete functionality of ART remains to be determined.

Anthropic clarified that its research initiative operates independently from pharmaceutical development. The organization confirmed it won’t pursue clinical trials related to this research.

Kauderer-Abrams informed Reuters that Anthropic doesn’t position itself as a competitor to pharmaceutical and biotechnology corporations. He explained that such companies concentrate on commercializing therapeutics, which falls outside Anthropic’s mission.

The organization maintains partnerships with pharmaceutical entities including Roche, Genentech, and Bristol Myers Squibb. While these corporations may utilize Anthropic’s AI capabilities, the company’s proprietary laboratory possesses resources unavailable to external academic researchers.

Certain researchers have expressed funding-related concerns. The Scientist publication reported in July that academics face difficult choices between allocating funds for AI computational tokens versus supporting graduate students and postdocs.

The ART discovery involved six identified human scientists collaborating with Claude. The initiative consumed 210 million tokens throughout the search and analytical phases.

Additional companies are pursuing AI-enhanced biological research. Both Alphabet’s Isomorphic Labs and OpenAI have released research findings in this domain during recent months.