An AI ran a mouse-brain injection. The stop button mattered.
This was one test, not proof that surgery is safe. WIRED reports a human-approved plan, a stop button, and a separate fix after a location error. No one has yet measured whether the drug helped.
What the source reports
A veterinarian first opened the mouse’s skull, WIRED reports. A person approved the plan before an AI system picked and carried out the injection. Staff had a button that could cut power to the machine. One model refused to do the live procedure, so the team switched models. A mismatch in how the system mapped locations then needed a separate repair step. The company account covers one mouse, not a safety trial. The effect of the drug had not been measured. A stop control and human approval are vital limits, but they do not show that this method is ready for routine care.
Original source
- Title
- An AI Agent Injected Drugs Into a Live Mouse’s Brain
- Author
- Emily Mullin
- Publication
- WIRED
- Date
- Saturday, October 31, 2026
Also covered by the other newsletter.