Anthropic has confirmed it’s running a physical biology lab in the San Francisco Bay Area — and this week said the lab has already made what the company calls a significant discovery: a previously unknown enzyme system with properties “reminiscent of CRISPR,” the gene-editing technology that transformed modern biology.
What Anthropic Actually Found
According to Anthropic, researchers at the lab identified a new enzyme system hidden in the DNA of bacteriophages — viruses that infect and replicate inside bacteria. CRISPR itself originated the same way: it’s a natural immune system bacteria use to fight off viruses, which scientists later adapted into the gene-editing tool now used across medicine and research worldwide. Anthropic hasn’t published full technical details of the new system publicly, and it’s too early to say whether it will prove as transformative as CRISPR did — but the comparison alone has drawn significant attention within the scientific and AI communities.
How the Lab Came to Exist
Reuters first reported the lab’s existence on September 18, citing two sources familiar with the matter, before Anthropic’s head of life sciences, Eric Kauderer-Abrams, confirmed it directly. “We believe that to do biology, the final test is still and will be for a while in real lab work,” he told Reuters. “We absolutely are doing that today.” The lab was set up quietly, without an official press release — the news surfaced through reporting rather than a company announcement.
Anthropic has been notably tight-lipped about the lab’s specifics: it hasn’t disclosed the facility’s size, staffing levels, opening date, or — notably — its biosafety classification, a rating that would indicate what kind of biological materials and risk levels the space is actually equipped to handle safely. A company spokesperson told Reuters the lab is “not for drug discovery specifically,” but declined to elaborate further on its actual purpose.
Humans, Not Claude, Ran the Physical Experiments
One detail stands out as particularly important for understanding the current risk picture: while Anthropic’s Claude models were involved in the broader research process, the physical lab experiments themselves were carried out by human researchers, not by AI systems operating lab equipment autonomously. Some reporting has framed this as arguably the most significant detail in the whole story — not the discovery itself, but the fact that Anthropic has, so far, kept a human in direct physical control of the experimental work.
Part of a Bigger Life Sciences Push
| Initiative | Detail |
|---|---|
| Claude Science | Drug discovery program launched June 2026, targeting rare diseases pharma isn’t prioritizing |
| Novo Nordisk partnership | Announced two days before the lab news, testing Claude Science tools on real drug-discovery problems |
| Life Sciences Verification Program | Launched the same week, giving vetted bio researchers access to Anthropic’s most powerful models |
| Model Hardware Standard | Released August 2026, a technical standard helping AI systems operate physical lab equipment |
| Pharma partners | Includes Roche’s Genentech, Bristol Myers Squibb, and Novo Nordisk |
Anthropic has said its wet lab work focuses on fundamental biology rather than drug discovery specifically, partly to avoid appearing to compete directly with pharmaceutical companies that are also its customers and partners — a sensitivity the company has run into before after launching products seen as competing with its own clients.
The Uncomfortable Timing
This news lands at a genuinely awkward moment for Anthropic. It comes just weeks after CEO Dario Amodei publicly called for the AI industry to slow down development, warning that models have grown capable — and potentially dangerous — enough to require far more careful safety testing. That call followed public warnings from two Anthropic researchers that advanced AI carries real, non-trivial risks, including one researcher’s estimate of a meaningful chance AI could pose an existential threat within the next decade.
Amodei himself has specifically named bioterrorism as one of his biggest fears about advanced AI — while also writing that he believes AI could “cure most major diseases in the next 5-10 years.” Anthropic has separately published reports acknowledging real situations where its own models may have been used to help someone attempt to develop biological weapons, as part of its ongoing safety and misuse tracking work. Taken together, this means the same company warning the world about AI’s bioweapon potential is now the same company running a physical biology lab whose safety classification and full research scope it hasn’t publicly disclosed.
Reuters Notes a Personal Dimension
Part of Anthropic’s broader biology push has reportedly been linked to Dario Amodei personally — his father died of a disease before a cure was found, a detail Reuters connected to the company’s stated motivation for pursuing rare-disease drug research through Claude Science and related programs.
What Remains Unclear
- The lab’s exact size, staffing, and precise opening date
- Its official biosafety classification level
- Full technical details of the newly discovered enzyme system, and whether it will prove genuinely useful the way CRISPR did
- The specific diseases or research areas the lab is ultimately working toward
- How much autonomy AI systems will be given over physical lab work as this program develops further
A Transparency Note
This story concerns Anthropic, the company behind the AI writing this article. We’ve aimed to report it factually and completely — including the genuinely unresolved tension between Anthropic’s public safety warnings and its undisclosed biosafety practices at this lab — rather than presenting only the discovery as a straightforward win.
Final Thoughts
A CRISPR-like discovery from an AI company’s own biology lab is a genuinely notable scientific claim — but the lack of disclosed biosafety information, combined with the timing right alongside Anthropic’s own public warnings about AI-assisted bioweapon risk, means this story raises as many open questions as it answers. Whether this becomes a case study in responsible, transparent AI-driven biology research, or a cautionary tale about disclosure gaps in a field where the stakes are unusually high, will likely depend on how much more Anthropic is willing to reveal about the lab going forward.
