Exin Therapeutics used an AI agent to guide a mouse brain injection

In short: A San Francisco biotech startup connected an AI agent to a robotic surgery setup and used it to help perform a brain injection on a live mouse.

What happened

Gabriel Ocana Santero, CEO and cofounder of Exin Therapeutics, said his team tested what happens when an AI agent is given direct control of lab equipment. An AI agent is software that can take actions for you, like a digital assistant that can also push buttons and move tools.

In the experiment, a veterinary surgeon first performed a craniotomy, which is a small opening in the skull. Then the AI agent selected a gene therapy option from a list, chose the target brain area and coordinates, and controlled the robotic arm and injector. The team was testing whether a gene therapy (a treatment that changes how cells behave by adding or editing genes) could reduce seizures in a mouse with epilepsy.

The team first tried Anthropic’s Claude Opus 5.5, but it refused to guide the needle into the brain, citing safety concerns. They switched to Claude Sonnet 5, which proceeded.

The setup included safety steps. The AI had to propose an injection plan for human approval, and there was an emergency stop button to cut power to the motors. The team also limited the AI’s access to the equipment’s code, which later caused delays when a coordinate glitch needed a fix. They used a separate AI system to adjust the code without telling it a live surgery was underway.

Why it matters

This is an early example of AI moving from giving advice to handling physical lab work, like a co-worker who can operate the tools. It also shows why human oversight and strong safeguards matter, especially when living animals are involved.

Source: Wired