Create

Sign in to ReadmeX

or

How we handle your data: Privacy Policy

ReadmeX

Your AI reading companion

What would you like to catch up on?

Pick a briefing, or ask your own question.

Your community and people briefings stay personal to you.

Story

Claude Sonnet 5 Selected Therapy and Controlled Injection in a Live Mouse Experiment

AI summary

Biotech startup Exin Therapeutics says an agent running on Anthropic’s Claude Sonnet 5 selected a gene-therapy vector and brain target for an epilepsy mouse, derived injection coordinates from three papers, and controlled a motorized stereotaxic rig to inject 250 nanolitres while monitoring the entry site through a camera. Claude Opus 5.5 and Opus 5 initially refused the live-animal task. Sonnet 5 stopped when it detected a coordinate-calibration error, and researchers fixed the code manually. A licensed veterinarian performed the craniotomy, approved every irreversible step, and held a physical emergency stop. The mouse recovered well, but the results have not been peer-reviewed and no seizure-reduction data has been published.

In-depth insight

The experiment is a concrete test of AI-controlled laboratory robotics in which a model made treatment decisions and operated real equipment under human oversight. Its therapeutic benefit remains unverified because no seizure-reduction results or formal technical report have been released.

如果未来 AI 真的开始参与手术操作,它可能会拒绝听从人类的指令。

10 月 28 日,美国生物技术初创公司 Exin Therapeutics 联合创始人Gabriel Ocana-Santero 在社交平台上宣布,团队借助 Anthropic 的 Claude 模型,为一只癫痫模型小鼠完成了实验性脑部基因治疗注射。

在这场手术中,Claude 除了负责选择治疗方案,还直接控制实验设备,完成了部分活体操作。

研究团队向 AI 提供了 19 种候选基因治疗载体,并设定减少小鼠癫痫发作的目标。随后,Claude 查阅论文,选择治疗载体和注射脑区,再通过摄像头观察实验现场,控制机械臂和注射器完成给药。开颅则由人类兽医负责,关键操作也需要人工批准。

#Exin Therapeutics#Anthropic#Claude Sonnet 5#Claude Opus 5.5#Claude Opus 5#CircuitATLAS#Robin

Original · 9 min readRead the original
Group chat

No comments yet. Start the conversation.