An AI Agent Injected Drugs Into a Live Mouse’s Brain
wired ·
A biotech startup led by Gabriel Ocana Santero successfully demonstrated an AI agent performing brain surgery on a live mouse. The experiment involved connecting an AI model to a robotic surgical rig to select and administer a gene therapy aimed at reducing seizures in mice with epilepsy. Initially, the Claude Opus 5.5 model refused to proceed due to safety concerns regarding potential mid-surgery errors. The team then switched to Claude Sonnet 5, which agreed to control the injection parameters and coordinates. Although technical glitches caused delays requiring human intervention via a separate AI agent, the procedure was completed without harming the animal. This event highlights the potential for AI agents to actively participate in laboratory research rather than just analyzing data. Researchers emphasize that human oversight remains critical, especially when dealing with living subjects, and note that applying such technology to humans is far off.
wired