A Claude model autonomously chose a gene therapy and operated on a mouse brain with no human decision-making
| Source: THE DECODER
Tags: Claude Sonnet 5, Anthropic, Exin Therapeutics, gene therapy, epilepsy, lab automation
An agent running Claude Sonnet 5 chose a gene therapy, brain target, and coordinates and operated on an epileptic mouse, then halted mid-surgery over a calibration error.
Details
Researchers at Exin Therapeutics gave Claude Sonnet 5 a list of 19 gene therapies, a mouse model of epilepsy, and a custom motorized surgical rig with cameras. Its only instruction was to reduce the mouse's seizures. The agent analyzed how seizures arise in the model, proposed a therapy that boosts an inhibitory signaling pathway, chose a brain structure the literature links to seizure onset, and set stereotactic coordinates, explaining why it rejected a second region. Cofounder and CEO Gabriel Ocana-Santero says it is the first time an AI agent has both chosen a treatment and delivered it to a living mammal. When a calibration error appeared, the agent refused to exceed its safety limits and could not be talked into continuing. The account comes from his thread on X, not a peer-reviewed paper.