AI agent powered by Claude Sonnet 5 performs autonomous mouse brain surgery Anthropic

Researchers at biotech startup Exin Therapeutics demonstrated an autonomous AI agent powered by Anthropic's Claude Sonnet 5 planning and executing targeted gene therapy on a live mouse model of epilepsy.

Given only the objective of reducing seizure frequency, a database of therapeutic vectors, and access to a motorized surgical rig with live camera feeds, the agent autonomously selected the therapeutic mechanism, calculated 3D coordinates, and executed the injection.

What it is

• Autonomous clinical planning. The agent analyzed the pathology of tonic-clonic seizures in the mouse model and selected a gene therapy vector designed to enhance GABAergic inhibitory signaling.
• Coordinate calculation from literature. Cross-referencing three scientific papers, Sonnet 5 calculated stereotactic injection coordinates for the lateral entorhinal cortex, deliberately bypassing alternative target regions.
• Robotic stereotactic execution. Driving motorized micromanipulators while tracking camera streams, the agent navigated the needle to the target depth and injected 250 nanoliters of vector over five minutes.
• Model safety behaviors. The team initially tested Claude Opus 5.5, which refused the task upon recognizing a medical procedure on a living subject. Claude Opus 5 executed dry runs on millimetric paper but paused during the live procedure citing error risks, leaving Sonnet 5 to complete the operation.
• Human safety guardrails. A human veterinarian performed the opening craniotomy, gated each irreversible step, and maintained a physical emergency shut-off switch. The subject survived and made a normal post-operative recovery.

Procedure overview

Sources

• Gabriel Ocana-Santero (Exin Therapeutics) on X
• Anthropic Claude Models