Gabriel Ocana Santero, CEO and cofounder of Exin Therapeutics, reported successfully using an AI agent to perform a brain injection procedure on a living mouse. The experiment involved connecting an AI agent to a custom-built surgical setup, which included a motorized device to immobilize the mouse and a camera for observation. The team tasked the AI model with selecting a treatment from 19 options in a spreadsheet, determining the appropriate brain region, and setting injection coordinates.
A veterinary surgeon performed the initial craniotomy while the AI agent controlled the robotic arms and injector needle, managing parameters such as volume and speed. The initial model tested was Anthropic’s Claude Opus 5.5, which reportedly refused to guide the needle into the mouse’s brain, suggesting it might stop midway through the surgery. The team switched to Anthropic’s Claude Sonnet 5, which proceeded with the procedure.
Ocana-Santero stated that the agent was able to choose the therapy and brain target and then deliver it into a living mammal, noting, “Nobody told this agent what to give or where to put it. It got a phenotype and a shelf of candidates and had to work out the therapy and the coordinates itself.” The procedure required several safeguards, including the AI proposing an injection plan for human approval and the presence of an emergency stop button.
During the operation, a technical glitch disrupted the equipment's coordinate system; although the agent recognized the calculated position was wrong, it refused to modify the code to correct the problem because its instructions prohibited it. The team resolved this by using a separate AI agent on Opus 5 to make the necessary code changes, though they did not inform this agent that a live surgery was occurring. The experiment lasted slightly longer than an hour, compared to the typical 45-minute procedure.
Anthropic stated that Opus is a heavy-reasoning intelligence specialist best for complex multistep problem-solving, while Sonnet is better suited for everyday tasks.
