US startup reports equipment fault during AI-controlled mouse experiment
US startup Exin Therapeutics connected an AI agent to laboratory equipment in a mouse experiment. Cofounder Gabriel Ocana Santero describes a malfunction that required intervention, with human approval and an emergency stop retained. His account concerns exploratory research under supervision; the team was still waiting to assess effects on seizures.
Artificial Intelligence··Morning
A founder describes a supervised experiment
Exin Therapeutics, a US startup that studies brain activity in its search for drugs against epilepsy and neurodegenerative disease, connected an AI agent to equipment during a live-mouse experiment. Cofounder Gabriel Ocana Santero described the work in an October 31 interview. His team was exploring whether agents could participate directly in laboratory research alongside human scientists, extending their use beyond experiment planning and analysis of results.[1]
An equipment fault interrupts the work
The founder reported a technical malfunction that interrupted the work and required a separate intervention. Human approval was part of the supervision, and researchers retained an emergency stop. Santero said the mouse was recovering well. The team was waiting to assess effects on seizures; Santero’s update on recovery preceded that outcome assessment.[1]
Laboratory observations motivate the research
Santero argues that published literature misses failed experiments, unexpected observations and decisions made as research unfolds. He wants agents to encounter that part of laboratory work, while retaining human oversight for animal experiments. His team plans further comparisons between models and safety checks before additional live-animal work. Anthropic did not provide a comment for the interview.[1]
