An AI agent carried out a brain injection on a live mouse after one Claude model refused, Wired reports

What happened

Wired reported that Gabriel Ocana Santero, CEO and cofounder of San Francisco-based startup Exin Therapeutics, used an AI agent to carry out a brain injection procedure on a living mouse. The procedure needed some extra steps, according to Wired, but it offers a look at how AI could take a more active role in laboratory research. Exin Therapeutics uses AI to analyze brain activity to develop drugs for epilepsy and neurodegenerative diseases.

Santero posted about the experiment on social media and told Wired he wanted to explore what happens when AI agents get direct access to lab equipment, instead of only designing experiments or analyzing results. "We've realized that we need to connect these agents to a lab to get the maximum value from them," he told the outlet.

Why connect AI to a lab

AI models are trained mostly on published scientific literature, Wired noted, which can leave out failed experiments, unexpected observations and decisions scientists make in real time. Santero said linking models to real-world experiments can make them more reliable.

How the procedure worked

The surgical rig

The team connected an agent to a custom-built setup used for brain injections in mice. According to Wired, the rig included a motorized device to hold the mouse still and a camera that let the agent watch the procedure, move an injection needle and control delivery of a compound. The experiment tested whether a gene therapy could reduce seizures in a mouse with epilepsy.

What the agent decided

The team asked the AI model to pick a treatment from 19 options in a spreadsheet. A veterinary surgeon first performed a craniotomy on the mouse while the agent observed. The agent then chose the therapy, the brain region and the injection coordinates, and it controlled the robotic arms, the injector needle and settings such as volume and speed.

One model said no

The team first tested Anthropic's Claude Opus 5.5, which refused to guide the needle into the mouse's brain and suggested it might stop midway through the surgery, which could endanger the animal. Santero told Wired the agent did the planning, but when it was time to inject it said, "something on my output is telling me I shouldn't do this." The team switched to Claude Sonnet 5, which agreed to proceed. Anthropic did not respond to Wired's request for comment.

Safeguards and a glitch

Wired reported the researchers required the AI to propose an injection plan for human approval and installed an emergency button a person could press to disconnect the rig's motors. They also restricted the agent's access to the equipment's underlying code so it could not modify its own operating rules.

A technical glitch then disrupted the equipment's coordinate system. The agent recognized that the calculated injection position was wrong but refused to change the code because its instructions explicitly prohibited that. The team used a separate agent running on Opus 5 to make the code changes, without telling that agent a live surgery was happening.

The back and forth delayed the procedure, Santero said. A typical brain injection surgery like this takes about 45 minutes, while this experiment lasted slightly longer than an hour.

The bigger picture

Wired noted that frontier AI labs have been planning for AI to take on more physical roles. In August, Anthropic released a framework for how its models can interact with the real world, and AI engineers more recently put GPT, Claude and Grok in charge of a car to pick up lunch at In-N-Out Burger. Only one of the models succeeded. Readers can find related coverage in our AI and tech section.

Santero told Wired he expects AI agents to increasingly work alongside human scientists with direct access to lab equipment. For now, he said, human oversight is key, especially when experiments involve living animals. Wired added that it is far too soon to turn an AI agent loose on a human brain.

What to watch next

• The team is waiting for the gene therapy to take effect before assessing its impact on the mouse's seizures.
• Exin Therapeutics plans to test Grok and other models next to see whether one performs better.
• The company is experimenting with agent-controlled craniotomies but wants to be sure of their safety before moving to live surgeries.