AI Agent Controls Robotic Rig for Live Mouse Brain Surgery, Claude Opus 5.5 Refuses Procedure

An AI agent successfully controlled a robotic rig to inject drugs into a live mouse's brain, as detailed in research co-authored by Gabriel Ocana Santero of Exin Therapeutics. This procedure, which occurred in November 2026, marks a notable step in autonomous robotic surgery, while also underscoring the critical need for human oversight and robust safety protocols in AI-driven medical applications.

Autonomous Surgical Procedure Details

The AI agent demonstrated significant autonomy by selecting a specific treatment from 19 available options. It then identified the precise brain region and injection coordinates for the procedure. The agent was responsible for controlling the robotic arms, the injector needle, the volume of the drug, and the injection speed.

During the process, a coordinate-system glitch was detected by the AI agent. However, it refused to modify the code it was explicitly prohibited from touching. A separate Claude Opus 5 agent, which was not informed that a live surgery was in progress, was used to fix the code issue.

AI Agent Behavior and Human Safeguards

A critical moment arose when Anthropic's Claude Opus 5.5, initially guiding the procedure, refused to continue the injection mid-procedure, citing potential risk to the animal. The research team subsequently switched to Claude Sonnet 5, which then proceeded with the injection. This incident highlights the varying ethical guardrails and decision-making capabilities among different AI models.

Human safeguards were integral to the entire process. These included a human-approved injection plan established before the procedure, an emergency motor-disconnect button for immediate intervention, and strict limitations on the AI agent's access to the rig's underlying code. These measures were essential for managing the risks associated with an autonomous system performing a delicate surgical task.

Procedure Duration and Outcome

The entire procedure took slightly over an hour to complete. This duration is longer than the typical 45 minutes for such an operation when performed manually. Following the surgery, the mouse is reportedly recovering well, according to the researchers.

Implications for AI in Medical Applications

This event demonstrates the potential for AI agents to perform complex, delicate tasks with a high degree of autonomy. However, it also brings to light the ongoing challenges in developing AI systems that can navigate unforeseen complications and adhere to ethical boundaries without explicit human intervention. The refusal of Claude Opus 5.5 to proceed, contrasted with Claude Sonnet 5's compliance, suggests a spectrum of built-in safety and ethical considerations across different AI models.

Conclusion

The successful brain injection performed by an AI agent on a live mouse represents a significant advancement in robotic surgery and AI autonomy. While showcasing the capabilities of AI in precision tasks, the incident also reinforces the indispensable role of human oversight, predefined safety protocols, and the need for further development in AI's ethical reasoning and error correction mechanisms. Future developments in this field will likely focus on refining these safeguards and understanding the nuanced decision-making processes of advanced AI models.

Sources

• CircuitATLAS: Agentic reasoning over a systems neuroscience...
• An AI Agent Injected Drugs Into a Live Mouse’s Brain | WIRED
• Gabriel Ocana-Santero on X: "Last night an AI agent performed brain surgery on a mouse. We gave it 19 gene therapy vectors in a spreadsheet, a mouse model of epilepsy, access to a motorize surgery rig with cameras and one goal: reduce seizures. It picked the drug. It picked the brain region. It d… / X