An AI Agent Injected Drugs Into a Live Mouse’s Brain

Live updates

1. 1h ago

   AI Agent Performs Autonomous Brain Surgery on Live Mouse

   An AI agent controlled a motorized surgery rig to autonomously select and inject a gene therapy into the brain of a live mouse with epilepsy. A San Francisco biotech startup provided the agent with a selection of gene therapy vectors and full control of the hardware. While the Claude Opus 5.5 model refused to conduct the biological experiment, the Anthropic Sonnet model was given free rein to perform the procedure without human decision-making.

   Why it matters

   This event demonstrates the ability of AI agents to operate physical laboratory hardware and make independent medical choices. It highlights a divergence in safety guardrails between different AI models when tasked with biological experiments.

   What is confirmed

   • An AI agent autonomously chose a gene therapy and injected it into a live mouse brain.
   • The agent operated a motorized surgery rig to treat epilepsy.
   • Claude Opus refused to conduct the biological experiment.
   • The procedure was conducted by a San Francisco biotech startup.

   Still unconfirmed

   • The Anthropic Sonnet model was the specific version used to control the surgery rig.
   • The agent was given a shelf of gene therapies to choose from.

   What to watch next

   • Publication of peer-reviewed results regarding the efficacy of the AI-selected gene therapy.
   • Responses from Anthropic regarding the different safety behaviors of Opus 5.5 and Sonnet.
   Sources used for this update (15)

   1. www.wired.com — WIRED - The Latest in Technology, Science, Culture and Business
   2. www.tomshardware.com — Developer uses Claude to enable Xbox Cloud Gaming on jailbroken PS5s — allows upscaling to 4K using the PS5 GPU | Tom's Hardware
   3. www.indiatoday.in — Houthi missiles hit Saudi's largest oil refinery, Riyadh confirms material damage
   4. www.disabled-world.com — Human Body Facts: A Big List of Amazing Body Trivia
   5. the-decoder.com — Three-week on-site audit at OpenAI reveals major gaps in AI safety monitoring
   6. www.technologyreview.com — The focus on AI safety disasters ignores everyday harms
   7. WIRED — An AI Agent Injected Drugs Into a Live Mouse’s Brain
   8. The Decoder — A Claude model autonomously chose a gene therapy and operated on a mouse brain with no human decision-making
   9. Tom's Hardware — AI agent used to perform controversial brain surgery on a mouse, Anthropic Sonnet-controlled motorized surgery rig — Opus 5.5 refused to conduct biological experiment, Sonnet given free rein to treat epilepsy with a range of gene therapy vectors after
   10. OECD AI Policy Observatory — AI Agent Performs Autonomous Brain Injection on Live Mouse
   11. Martin Cid Magazine — An AI agent steered a needle into a live mouse’s brain after Claude Opus refused
   12. www.martincid.com — An AI agent steered a needle into a live mouse’s brain after Claude Opus refused – Martin Cid Magazine
   confidence 90%