Claude-based agent guided a mouse brain surgery test
BROBROK News · Based on reporting by DiarioBitcoin ·
A Claude Sonnet 5-based agent reportedly chose a gene therapy, the brain target, and the injection settings for an experiment in a mouse with seizures. Mid-procedure, it stopped after flagging that the target was outside the device’s physical range, and the team later fixed the issue manually.
Key points
- The agent compared 19 candidate gene therapies before selecting one for the test.
- It also chose the brain target, stereotactic coordinates, and injection parameters from scientific papers.
- A licensed veterinarian approved each irreversible step and kept a physical emergency stop available.
- The system refused to continue when it detected a mismatch between the target and the machine’s range.
- The results have not been peer-reviewed yet.
Why it matters
The demonstration suggests AI agents may be able to take part in experimental planning and device control in a live biological setting. It also shows that automated safety checks can block a risky move, but human oversight is still needed to diagnose and resolve the underlying problem.
Original headline: “Claude eligió una terapia génica y operó el cerebro de un ratón, pero se detuvo por seguridad”
Read the original on DiarioBitcoin
Summary and analysis written by BROBROK News from the original report. Read the full story on the source website. Not investment advice.