# Anthropic Built a Wet Lab in the Bay Area and Is Running Physical Biology Experiments

Life sciences head Eric Kauderer-Abrams said the company is exploring having Claude direct laboratory robots, while a spokesperson said the lab is not dedicated to drug discovery and will run no clinical trials.

- Published: 2026-09-19T06:16:00.678Z
- Canonical: https://polylog.news/ai/2026-09-19/anthropic-built-a-wet-lab-in-the-bay-area-and-is-running-phy
- Publisher: Polylog (AI desk)
- Section: tech
- Sources: [Polylog editors](https://polylog.news), [Gizmodo](https://gizmodo.com/while-issuing-warnings-about-ai-made-bioweapons-anthropic-quietly-set-up-an-ai-run-robot-operated-biology-lab-here-in-the-us-2000814092), [Yahoo Finance](https://finance.yahoo.com/healthcare/articles/anthropic-builds-wet-lab-ai-130141303.html)

Anthropic has set up a physical biology laboratory in the San Francisco Bay Area and is already running experiments there, the company's head of life sciences, Eric Kauderer-Abrams, confirmed this week. The plan, as [reported by Gizmodo](https://gizmodo.com/while-issuing-warnings-about-ai-made-bioweapons-anthropic-quietly-set-up-an-ai-run-robot-operated-biology-lab-here-in-the-us-2000814092) and [summarized by AI Post](https://t.me/aipost/8190), is to pair Claude with laboratory automation so robotic systems execute experiments with limited human intervention.

Kauderer-Abrams described a hybrid arrangement that matches normal biotechnology practice, splitting work between Anthropic's own benches and outside partners, and said the team is early in automating lab execution. A company spokesperson added two limits: the lab is not dedicated to drug discovery, and Anthropic will not run clinical trials. Reporting on the effort frames the target as research into rare and hard-to-treat conditions, including diseases with no tractable drug target, as [Yahoo Finance noted](https://finance.yahoo.com/healthcare/articles/anthropic-builds-wet-lab-ai-130141303.html).

The move is in tension with Anthropic's own public position. The company has argued more forcefully than any other lab that frontier models raise biological misuse risk, and its September threat report described blocking attempts to use Claude for work relevant to biological weapons. Building an in-house facility where a model directs physical biology work means Anthropic now holds both roles, the party warning about the capability and the party extending it.

Nothing published so far shows a model designing and executing a novel experiment autonomously. What exists is a facility, an intention, and a claim that early automation is underway. The test is whether Anthropic publishes a closed-loop result where a model proposed a hypothesis, ran the assay, and revised based on the measurement.

## What this means

Closed-loop laboratory automation moves the competitive frontier for frontier labs from token generation to experimental throughput, which favors whoever controls both the model and the instruments and pressures contract research organizations that sell only execution. It also changes the safety argument: a lab that runs its own wet bench can no longer treat biological risk as something that happens on someone else's premises, and regulators reviewing biosecurity controls now have a US facility to inspect rather than an abstraction.

## What to watch

- Whether Anthropic publishes a peer-reviewed result produced by the closed loop, which is the difference between a research program and a press position.
- Whether any biosecurity regulator or Congressional committee asks for oversight terms on a model-directed laboratory, since that would be the first governance test of this arrangement.
- Whether laboratory automation vendors announce Anthropic integrations, which would show the agent-to-instrument interface is being standardized rather than built in-house.