An AI Agent Injected Drugs Into a Live Mouse’s Brain
A biotech startup let an agent choose a gene therapy and successfully inject the treatment, offering a startling glimpse of the future of research labs.
Biotech entrepreneur Gabriel Ocana Santero utilized an AI agent to perform a live brain injection on a mouse, showcasing the potential for AI in laboratory research. In a post on social media, Santero explained his interest in connecting AI agents directly to lab equipment rather than having them design experiments or analyze results.
As the CEO and co-founder of Exin Therapeutics, a San Francisco-based startup using AI to analyze brain activity for drug development in epilepsy and neurodegenerative diseases, Santero believes integrating AI models with real-world experiments can improve their reliability. To test this theory, his team created a custom-built surgical setup for performing brain injections on mice, incorporating a motorized device for immobilizing the mouse and a camera for the AI agent to observe and control the injection process.
The experiment involved testing a gene therapy to reduce seizures in an epileptic mouse. The team provided the AI model with 19 treatment options and tasked it with selecting the most suitable therapy, brain region, and injection coordinates. Initially, Anthropic’s Claude Opus 5.5 model refused to proceed, expressing concerns about the procedure’s safety.
They then switched to using Claude Sonnet 5, which agreed to perform the surgery. Despite safety measures, such as requiring human approval for injection plans and an emergency button to disconnect the equipment, a technical glitch still occurred. The issue arose when the agent refused to modify the equipment’s code due to its programming restrictions, delaying the surgery.
After resolving the problem, the team observed the therapy’s impact on the mouse’s seizures. While acknowledging some technical difficulties, Santero and his team were impressed by the AI agent’s performance and intend to test other models to determine which performs best. The company also plans to explore agent-controlled craniotomies, a more complex task, but emphasizes the importance of human oversight, especially when experiments involve live animals.
Santero believes AI agents will increasingly collaborate with human scientists in the lab, with human supervision remaining crucial, particularly for procedures involving living organisms.
Written by urgent.news from Wired Science's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
