AI Agent Performs Brain Surgery on Live Mouse
Original: An AI Agent Injected Drugs Into a Live Mouse’s Brain
Why This Matters
Marks a shift from AI as lab assistant to AI as active physical researcher.
Exin Therapeutics CEO Gabriel Ocana Santero used an AI agent to select a gene therapy and inject it into a living mouse's brain, controlling robotic arms, needle placement, and injection parameters in a first-of-its-kind lab experiment.
San Francisco biotech startup Exin Therapeutics has demonstrated an AI agent performing a live brain injection on a mouse — selecting the treatment, targeting coordinates, and controlling the robotic hardware end-to-end. The company, which uses AI to analyze brain activity for epilepsy and neurodegenerative drug development, built a custom surgical rig with a motorized mouse restraint, camera, injector needle, and robotic arms, all controllable by an AI agent.
For the experiment, the agent chose from 19 gene therapy candidates in a spreadsheet, aiming to reduce seizures in an epileptic mouse. A human veterinary surgeon first performed a craniotomy; the agent then took over, selecting the therapy, brain region, and injection coordinates, and managing volume and speed parameters.
Notably, Anthropic's Claude Opus 5.5 refused to guide the needle mid-procedure, citing internal output signals suggesting it should stop — a refusal that could itself have harmed the animal. The team switched to Claude Sonnet 5, which completed the surgery. Santero says the goal is getting AI agents direct access to lab equipment rather than limiting them to experimental design or data analysis. 'We've realized that we need to connect these agents to a lab to get the maximum value from them,' he told WIRED. The approach could help models learn from real-time observations typically absent from published literature, including failed experiments and in-the-moment decisions.