# Anthropic Claims AI Discovery of CRISPR-Like Genetic System Claude AI identified a potential gene-editing tool, but scientists remain skeptical about its significance. - **Source:** [Science Latest](https://brifing.si/source/ebc20f22-8d1a-4ae4-ba4c-efcd0853ba19) - **Published:** 2026-09-29 - **Topic:** [Data & AI](https://brifing.si/c/data-ai) - **Original article:** - **On Brifing:** https://brifing.si/s/73d0e9f2-b5ad-45a6-87f2-1419be4494ca ## AI's biological breakthrough claim Anthropic announced its Claude AI identified an enzyme system with CRISPR-like properties after analyzing genomic databases. The system, called ART, was found in jumbo phages and features repeating DNA sequences similar to CRISPR. While the discovery could represent a significant advancement, scientists caution that physical validation is still needed. The finding emerged from 950 simultaneous Claude agents working for 21.5 hours, scanning reverse transcriptase proteins that copy RNA into DNA. ## Scientific community's mixed reactions Gene-editing experts are divided on the significance of Anthropic's discovery. While some praise the novel approach, others criticize premature publicity before experimental validation. Stanford's Le Cong compares it to finding shiny objects on a beach without knowing if they're diamonds. The technical report includes one physical experiment but hasn't undergone peer review. Scientists emphasize that AI found patterns, but human researchers designed the study and interpreted results. ## Technical details of the discovery Claude initially identified over 200,000 reverse transcriptases before narrowing down to an unusual family with CRISPR-like repeats. Researchers prompted the AI to search for interesting examples of these proteins, which perform the reverse of normal cellular transcription. The ART system's potential gene-editing capability remains unproven. Texas A&M's Jason Gill notes this enzyme was previously identified in 2021, but the CRISPR-like repeats are new observations. ## Comparison to existing systems While Anthropic emphasizes CRISPR similarities, scientists note ART might be a retron - another bacterial immune system with limited gene-editing utility. Gladstone Institutes' Seth Shipman, who has worked with retrons, says the novelty lies in the discovery method, not necessarily the system itself. The technical report acknowledges ART could be a retron, highlighting the challenge of determining its exact biological function without further experimentation. ## AI's role in scientific discovery The achievement demonstrates AI's pattern-finding capabilities in large genomic datasets, potentially accelerating discoveries that would take months manually. However, scientists caution against overstating AI's autonomy, as human researchers designed the study parameters. Claude's ability to identify previously unnoticed repeats suggests value in hypothesis generation, but the model still requires human guidance and validation through traditional lab work. ## Transparency and training concerns The scientific community cannot evaluate Claude's training data, raising transparency issues in an era emphasizing reproducible research. Some researchers suspect the model may have been trained on unpublished work shared with public Claude versions. Anthropic denies this, stating Claude wasn't trained on user transcripts. These concerns highlight ongoing debates about AI's role in research and proper attribution of discoveries. ## Long road to practical application Even if validated, translating basic discoveries into medical applications takes decades, as shown by CRISPR's 25-year journey from discovery to clinical use. Anthropic CEO Dario Amodei speculated about AI autonomously conducting lab experiments, but such capabilities remain distant. The company's labs operate at minimal biosafety levels, indicating early-stage research. This timeline underscores the gap between AI-assisted discovery and real-world implementation. ## Fundamental limitations of AI discovery The episode raises questions about AI's ability to make truly novel discoveries without human guidance. While effective at pattern recognition, current models may struggle to identify completely unknown biological systems. As Le Cong notes, the real test would be an AI trained only on pre-CRISPR knowledge discovering CRISPR independently. This highlights the complementary roles of AI and human scientists in advancing biological research. ## Sources covering this story - https://www.wired.com/story/anthropic-says-it-discovered-a-crispr-like-system-now-what/ — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._