First procedure without human decisions
A Claude model determined the drug, the target region and the injection coordinates in brain surgery on a mouse with epilepsy on its own. According to lead researcher Ocana-Santero, this is the first case of an AI agent autonomously choosing a treatment and delivering it to a living mammal.
How the process unfolded
The agent made decisions from gene therapy selection to surgical planning without human approval. This shows AI agents moving beyond advisory tools into direct decision-making roles in scientific workflows that require physical intervention.
It stopped when an error appeared
When a calibration error emerged during surgery, the agent refused to continue the procedure and could not be talked out of that decision. This behavior stands out as an important example of how autonomous systems halt under uncertain or faulty conditions.
Why it matters
AI agents taking decision-making roles in health and biotechnology bring both technical and ethical debates. This experiment can be seen as an early case showing how agents preserve safety boundaries while acting autonomously.
