# AI Watched a Live Brain Operation in Real Time and Helped Save a Man's Sight > Source: > Published: 2026-08-27 11:01:26+00:00 *A London surgical team has moved AI from the research bench into a live brain operation, using it to help protect a patient's sight while removing an 11mm pituitary tumour.* The stakes were brutally simple. Rhys Hibbert, 48, from Bedfordshire, had a tumour pressing close to the nerves that help control vision, and one bad move inside that part of the brain could have left him blind. In May 2026, surgeons at the National Hospital for Neurology and Neurosurgery in London removed it while an AI system watched the live surgical video feed and marked the structures they needed to avoid. That is the story. Not a chatbot in a hospital press office, not a promise about medicine years from now, but software built at University College London reading moving footage while the operation was actually happening. According to University College London Hospitals, which announced the case on 27 August 2026, it was the first time AI had supported a neurosurgeon in real time during live surgery on a patient. Hibbert's condition was found in December 2024 after he collapsed during a walk and had a seizure. Doctors later found a tumour measuring around 11mm on his pituitary gland, the marble-sized organ at the base of the brain. His symptoms worsened over time, including hormone problems and trouble with his vision. By the time surgery became the right option, he had also volunteered to take part in research. When he woke up, the result was plain. UCLH quoted Hibbert saying, "When I came round... I could see everything in the room clearly." Within a week, the hospital said, he was walking independently without glasses or sticks. You don't need to dress that up. For a patient who had been tripping because he couldn't see the lower field of vision, clear sight was the point. [Chinese State Hackers Doubled Their Attack Volume Using DeepSeek AI](https://startupfortune.com/chinese-state-hackers-doubled-their-attack-volume-using-deepseek-ai/) Chinese state-affiliated hacking groups have more than doubled their attack volume after integrating DeepSeek into their operations, according to Taiwanese research firm TeamT5. Separately, Palo Alto Networks' Unit 42 found an independent hacker used DeepSeek to autonomously scan and breach 14 of more than 460 targeted systems. - [chinese state hackers using deepseek AI for attacks](https://startupfortune.com/chinese-state-hackers-doubled-their-attack-volume-using-deepseek-ai/) - [how AI models are scaling cyberattack operations](https://startupfortune.com/chinese-state-hackers-doubled-their-attack-volume-using-deepseek-ai/) ## The AI watched, but the surgeons operated A millimetre is not much room. In this part of the brain, the pituitary gland, blood vessels and nerves controlling vision sit tightly together, and UCLH said going a millimetre wrong can make the difference between a good operation and death, blindness or stroke. Surgeons already work with scans, training and experience, but the body doesn't always behave like the scan once tissue starts moving. That distinction matters. Most medical AI tools read static material: a scan taken before surgery, a slide reviewed later, a dataset pulled into a system after the fact. This one analysed the live surgical video feed in real time and helped highlight critical anatomy at the base of the brain while the surgeon worked. The line around control is important here. The trial was funded by the National Institute for Health and Care Research and Google, and the hospital has been careful to say this was not robotic surgery. The AI did not hold an instrument. It did not make the cut. Hani Marcus, consultant neurosurgeon at the NHNN and professor of neurosurgery at UCL's Queen Square Institute of Neurology, performed the operation with Danyal Khan, the surgical resident leading the work. Frankly, that is the more useful version of AI in an operating theatre. Nobody needed a machine pretending to be a surgeon. They needed a system that could keep watching the same camera feed, identify nerves and vessels by colour coding, and give the human team another layer of warning before damage was done. ## The harder part starts after the first case UCLH had used the technology as a research and training tool before, but the May operation was the first time it was used on a live patient. The system was developed at the UCL Hawkes Institute and runs on an NVIDIA Clara IGX platform, which is designed for real-time AI in medical device settings. UCLH said researchers trained and evaluated it using a large collection of annotated endoscopic pituitary surgery videos from earlier operations. Dr Sophia Bano, the UCL robotics and AI academic who is technical lead for the system, said it had learned from hundreds of surgical videos. That matters because a surgeon might spend years building up that range of examples, while software can be exposed to many past cases before it ever enters theatre. Still, there is no magic in the phrase "hundreds of videos." The real test is whether the system can stay reliable when the view is messy, the anatomy is unusual, or the operation moves beyond a tightly defined pituitary case. So don't turn one successful operation into a medical revolution. A pituitary tumour resection is a sensible place to start because the anatomy is dangerous but well mapped. Emergency trauma, tumours in awkward places, and procedures with less predictable tissue movement are harder problems. UCLH says the technology may also track surgical instruments and tissue interactions in future, which is useful only if it keeps proving itself under clinical pressure. [How One Judge's Split Ruling on Anthropic Became AI's Copyright Rulebook](https://startupfortune.com/how-one-judges-split-ruling-on-anthropic-became-ais-copyright-rulebook/) Judge Alsup ruled that training AI on books is fair use, but that Anthropic's piracy of those books from shadow libraries was not, a split that cost the company $1.5 billion. Publishers are now using the same legal seam to sue Google over how Gemini was trained. - [how judges ruled on AI training copyrighted books](https://startupfortune.com/how-one-judges-split-ruling-on-anthropic-became-ais-copyright-rulebook/) - [why anthropic paid 1.5 billion for copyright infringement](https://startupfortune.com/how-one-judges-split-ruling-on-anthropic-became-ais-copyright-rulebook/) The cautious rollout is part of why this story is worth your time. The operation happened in May, and details were kept private until Hibbert had recovered, according to reports carried by The Guardian and PA-linked coverage. That is what you want from hospital AI: a working case, a named patient, a named surgical team, a clinical trial, and a delay long enough to know the patient came through it. AI in health care is usually sold too loudly. This case doesn't need that. A customer services manager and community volunteer from Bedfordshire went into brain surgery with his sight at risk, and came out able to see the room clearly. The next question is whether the same kind of live support can help more surgeons make fewer mistakes when there is no room for one. **Also read:** [Google DeepMind Is Losing the AI Talent War It Used to Win Easily](https://startupfortune.com/google-deepmind-is-losing-the-ai-talent-war-it-used-to-win-easily/) • [Salesforce Stock Jumps 14% as Benioff's AI Numbers Silence SaaS Doubters](https://startupfortune.com/salesforce-stock-jumps-14-as-benioffs-ai-numbers-silence-saas-doubters/) • [GLM-5.3-Flash Is China's Real Strike at American AI Dominance, Not Just a Model Release](https://startupfortune.com/glm-53-flash-is-chinas-real-strike-at-american-ai-dominance-not-just-a/)