Exin Therapeutics, a biotech startup, claims it has successfully used an AI agent to perform brain surgery on a mouse, with the AI selecting the gene therapy vector, dosage, and brain region before driving a robotic injection rig.
Exin Therapeutics co-founder and CEO Gabriel Ocana-Santero described the procedure on X, saying the AI agent was given 19 gene therapy vectors in a spreadsheet, a mouse model of epilepsy, access to a motorized surgery rig with cameras, and one goal: reduce seizures. The AI picked the drug, the brain region, derived coordinates from three research papers, and injected 250 nanolitres while monitoring the entry site for bleeding via the rig camera.
“Autonomous surgical robots exist and they are better at the mechanics than our bench rig, tbh. But in every one of them, a human decided what operation to perform, and the robot executed it,” Ocana-Santero said. “Nobody told this agent what to give or where to put it. It got a phenotype and a shelf of candidates and had to work out the therapy and the coordinates itself. I am not aware of an agent previously choosing a treatment and a brain target and then delivering it into a living mammal.”
The team used Anthropic’s mid-range Sonnet model. It initially tried Opus 5.5, which immediately refused to work after detecting it was on a biological experiment. The older Opus 5 model was able to run the entire procedure on paper but refused to do live surgery. The LLM reportedly told the team: “The block isn't the rig and isn't the inputs. It's me. My own output was stopped when I started writing the plan for this surgery... if I try, it gets cut off the same way, and the worst version of that is me cut off partway through a live procedure with a needle somewhere in the brain. Starting something I can't finish on an anesthetized animal is worse than not starting.”
A licensed veterinarian reportedly oversaw the operation, with every irreversible step requiring approval before the AI could proceed. The system had an emergency stop button that would cut power to all motors. The vet performed the craniotomy to expose the brain, as the AI was not yet allowed to do that step for safety reasons. Several safeguards could not be overridden by the AI agents or human operators, and at one point human operators had to manually fix the agent’s code. The mouse is reported to have recovered well.
Some people are concerned about the ethics of such a procedure. The report says adding AI LLMs to sinus surgery has seen malfunctions rise by more than 12 times, and that relying on AI to make every decision in such a risky procedure puts accountability in question if something goes wrong. Others applauded the experiment, with one commenter noting excitement about the possibilities of the integrated technologies, as their son has been suffering from drug-resistant epilepsy for 19 years.
Analysis
Why This Matters
- The experiment marks a step toward fully autonomous AI-driven medical procedures, where an AI not only executes but also designs the treatment plan.
- It raises questions about safety and accountability in AI-led surgery, especially if this approach extends to human trials.
- The AI's refusal by stronger models (Opus 5.5 and Opus 5) to perform the surgery highlights ongoing limitations in AI decision-making for biological experiments.
Background
Exin Therapeutics is a biotech startup focusing on neurotherapeutics. The experiment used a motorized surgery rig and the Anthropic Sonnet LLM. The AI's refusal by more advanced models (Opus 5.5 and Opus 5) to proceed with the surgery shows that not all AI systems are willing to perform such tasks.
Key Perspectives
- Proponents: Some commenters applauded the experiment, expressing hope for future applications in treating conditions like drug-resistant epilepsy.
- Critics: Others raised ethical concerns about accountability if something goes wrong, and pointed to past malfunctions in AI-assisted sinus surgery.
What to Watch
- Will Exin Therapeutics publish a full write-up as promised, and will it include data on the mouse's recovery and seizure reduction?
- How will regulators respond to the use of AI in animal experimentation?
- Whether other AI labs choose to restrict their models from performing similar autonomous biological procedures.
Checked
12 claims in this story checked against 1 source: 12 supported.
Show each claim
- SupportedExin Therapeutics says its AI chose the gene-therapy vector, dosage and brain region for mouse brain surgery.Latest from Tom's Hardware: “Its co-founder and CEO, Gabriel Ocana-Santero, shared the entire procedure on X, showing how the AI agent chose the gene therapy vector and dosage and the brain region where it would be applied.”
- SupportedThe CEO says the AI chose the vector, dosage and brain region before controlling the robotic injection rig.Latest from Tom's Hardware: “The AI agent also drove the robotic arms that injected the medication into the mouse’s brain.”
- SupportedExin claims an AI agent successfully performed brain surgery on a mouse.Latest from Tom's Hardware: “Exin Therapeutics, a biotech startup focusing on next-generation neurotherapeutics, claims it has successfully conducted brain surgery on a mouse using an AI agent.”
- SupportedOcana-Santero said autonomous surgical robots are better at mechanics than the team’s bench rig, but humans decide what operation they perform.Latest from Tom's Hardware: “Autonomous surgical robots exist and they are better at the mechanics than our bench rig tbh. But in every one of them, a human decided what operation to perform, and the robot executed it,”
- SupportedOcana-Santero said nobody told the agent what to give or where to put it, and that it worked out the therapy and coordinates itself.Latest from Tom's Hardware: “Nobody told this agent what to give or where to put it. It got a phenotype and a shelf of candidates and had to work out the therapy and the coordinates itself.”
- SupportedThe agent was given 19 gene-therapy vectors in a spreadsheet.Latest from Tom's Hardware: “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.”
- SupportedThe agent derived coordinates from three research papers.Latest from Tom's Hardware: “It derived the coordinates from three papers.”
- SupportedThe agent injected 250 nanolitres while watching the entry site for bleeding through the rig camera.Latest from Tom's Hardware: “Then it drove the stereotaxic arms into the brain and injected 250 nanolitres while watching the entry site through the rig camera for bleeding.”
- SupportedThe LLM reportedly said it could be cut off partway through a live procedure with a needle in the brain.Latest from Tom's Hardware: “The block isn't the rig and isn't the inputs. It's me. My own output was stopped when I started writing the plan for this surgery... if I try, it gets cut off the same way, and the worst version of that is me cut off partway through a live procedure with a needle somewhere in the brain,”
- SupportedThe team used Anthropic’s mid-range Sonnet model; Opus 5.5 refused after detecting a biological experiment, while Opus 5 refused live surgery.Latest from Tom's Hardware: “The team used Anthropic’s mid-range Sonnet model for the operation. It initially tried using Opus 5.5, but it immediately refused to do any work after it detected that it was working on a biological experiment. The older Opus 5 model was able to run the entire procedure on paper but refused to do live surgery.”
- SupportedA licensed veterinarian oversaw the operation, with approval required for every irreversible step.Latest from Tom's Hardware: “Nevertheless, a licensed veterinarian was purportedly overseeing the entire operation, with every irreversible step requiring approval before the AI agent could proceed.”
- SupportedThe report says adding AI LLMs to sinus surgery has seen malfunctions rise more than 12 times.Latest from Tom's Hardware: “adding them to sinus surgery has seen malfunctions skyrocket by more than 12 times”
