AIToday
Large Language ModelsAI Business & IndustryTHE DECODERPublished: Sep 25, 2026, 13:00 JST

Anthropic touts Claude's ART enzyme find; CRISPR researcher says routine genome mining

Anthropic touts Claude's ART enzyme find; CRISPR researcher says routine genome mining

3 Key Points

  1. What happened

    Anthropic said Claude largely on its own found an unknown enzyme system, "ART," in DNA databases. About 950 AI agents spent 21 hours scanning over 200,000 reverse transcriptases, with humans setting the question and running lab experiments.

  2. Why it matters

    The finding suggests AI could speed up early-stage genome mining, though the body does not confirm what ART does.

  3. What to watch

    CRISPR researcher Lucas Harrington says such methods date to decades ago and similar systems have been known since 2008. The test is whether Anthropic shows a function for ART, not just its existence in bacteriophages.

WHO IT HITSAI and biotech research teams that rely on genome mining to find new gene-editing tools may need to weigh AI-generated leads more carefully, since a named CRISPR researcher says the output so far is routine, not a proven discovery.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, in your inbox every morning.

Context & Analysis

Anthropic is not presenting the ART finding as a standalone paper but as one of the first discoveries from its Bay Area biology lab, which opened in spring 2026. The company says Claude did most of the analysis on its own, while humans set the research question and ran the lab experiments — a division of labor that places AI on the hypothesis-generating side of biology, not the bench.

The technical claim rests on scale: about 950 AI agents spent 21 hours combing through more than 200,000 reverse transcriptases. Next to one unusual enzyme, they found repeating DNA sequences. That is the signature of genome mining, a search method that Mammoth Biosciences, co-founded by Lucas Harrington, already uses to hunt for CRISPR systems.

Harrington's pushback is not that the search failed, but that finding a system is only the first step. The hard part is figuring out what a system actually does, and Anthropic has not shown that, he wrote on X. Whether ART becomes a meaningful result may hinge on whether Anthropic or others can demonstrate a function, rather than on how much of the search the AI ran.

FAQ
What is ART?
ART is a previously unknown enzyme system that Anthropic says Claude found in DNA databases. It shows up mostly in bacteriophages and resembles the gene-editing tool CRISPR, but its function is still unknown.
Did Claude do the work in the lab?
No. Humans set the research question and ran the lab experiments. About 950 AI agents spent 21 hours combing through more than 200,000 reverse transcriptases.
Why are CRISPR researchers skeptical?
Lucas Harrington, who earned his PhD under Nobel laureate Jennifer Doudna and co-founded Mammoth Biosciences, says the method has been around for decades and similar systems have been known since 2008.

Get the latest Large Language Models news every morning

For example, today's edition would include:

  • Schmidhuber joins Sakana AI to lead RSI LabITmedia AI+ · 1h ago
  • TypeSafe AI launches Jev, claiming up to 200x faster inference without hallucinationsITmedia AI+ · 1h ago
  • Sakana AI names Jürgen Schmidhuber Chief Scientific AdvisorTHE DECODER · 1h ago

AI-summarized, only the topics you pick: one digest a day via Email, LINE, or Slack.

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.

Questions and answers are published on this page.

Related Articles

Next articleOpenAI to preview GPT-6 Cyber in weeks, with new deployment tool