AI agent planned and guided a brain injection in a live mouse
The procedure ran longer than a typical surgery of this kind, and the mouse is reported to be recovering while its seizure response is assessed.
A biotech startup, Exin Therapeutics, let an AI agent run a brain injection on a live mouse with epilepsy, with a veterinarian performing the craniotomy first. The agent picked a gene therapy from 19 candidates and a target region, then drove the robotic arm and injector. The team first tried Claude Opus 5.5, which declined to guide the needle because it might stop partway through the surgery, so they switched to Claude Sonnet 5, which agreed. Safeguards included human approval of the plan and an emergency stop, and a coordinate fault had to be fixed by a separate agent on Opus 5 because the primary agent was barred from editing its own code. The procedure ran a little over an hour against a typical 45 minutes, and the mouse is reported to be recovering well while its seizure response is assessed.
Why it matters
The case shows an AI agent choosing the therapy and target and driving the robotic injector, with a vet doing the opening and humans approving each irreversible step. It also shows that one model declined the task while another agreed, so the outcome depended on which model was used. Readers following AI in lab work can see that the human approval steps were the points where the procedure was held to account.
Matters or hype?
Does this story matter, or is it hype? Vote before you see what everyone else said.