An AI Agent Injected Drugs Into a Live Mouse’s Brain

An AI agent controlled key parts of a brain injection procedure on a live mouse, including choosing a gene therapy and operating the injector. The experiment offers an early example of AI taking an active role in laboratory research, where it could learn from practical conditions that are often missing from published papers.

The team was testing whether one of 19 possible therapies could reduce seizures in a mouse with epilepsy. A veterinary surgeon first opened the skull, after which the agent selected the brain region and coordinates and controlled the injection; researchers required human approval of its plan and provided an emergency stop. An equipment glitch disrupted the coordinate system, and the agent would not alter its code because its instructions barred it, so a separate AI agent corrected the problem without being told live surgery was under way.