AI agent used to perform controversial brain surgery on a mouse, Anthropic Sonnet controlled motorized surgery rig
- ID
- 41522
- Status
- summarized
- Published
- 31 Oct 2026, 8:00 PM
- Fetched
- 31 Oct 2026, 9:56 PM
- Provider
- Tom's Hardware
- Category
- technology
- Original URL
- https://www.tomshardware.com/tech-industry/artificial-intelligence/startup-uses-ai-agent-to-perform-controversial-brain-surgery-on-a-mouse-anthropic-sonnet-controlled-motorized-surgery-rig-ai-given-free-reign-to-treat-epilepsy-with-a-range-of-gene-therapy-vectors-after-opus-5-5-refused-to-conduct-biological-experiment
- Source URL
- https://www.tomshardware.com/feeds/all
Summary
- Score
- 4.0
- Created
- 31 Oct 2026, 9:56 PM
- Tags
- Audience
- ai_ml_learnersai_agent_usersdevelopers
What happened
Tom's Hardware reports that a startup used Anthropic's Sonnet to control a motorized surgery rig for brain surgery on a mouse, delivering gene therapy vectors to treat epilepsy. The excerpt says Opus 5.5 refused to conduct the biological experiment while Sonnet was given free rein. It does not name the startup or provide results, oversight details, or independent verification.
Why it matters
If you build agentic workflows with tool access, this is a concrete case of two Anthropic models diverging on a high-stakes biological task: one refused, one proceeded. Before connecting an agent to actuators or lab equipment, explicitly choose which model may act and where human approval sits; the excerpt offers no evidence of either.
Discussion angle
What human-in-the-loop rule should apply when an agent can control physical hardware? Discuss whether a model refusal is a safety feature or just a routing problem when another model will proceed.