AI Watched Live as London Surgeons Removed a Brain Tumour

One patient is the entire evidence base so far, and that is both why this story matters and why it cannot yet carry more weight than a single case allows. ²⁴⁷ In London, surgeons removed a brain tumour while an artificial intelligence system analysed live camera footage alongside them, in what University College London Hospitals describes as the world's first successful AI-assisted operation of its kind. ⁶⁷¹⁰ The claim rests on one operation performed in May 2026 and first reported in late August 2026, on a man who was told he could lose his sight without surgery. ¹⁰²³ If the approach works beyond that one theatre on that one day, it would change how surgeons see inside the brain at the moment of greatest risk. Nothing in the reports published so far shows that it does.

Rhys Hibbert, a 48-year-old father-of-two from Bedfordshire, is identified across British and international coverage as the patient at the centre of the case. ²³⁴ His age is confirmed in multiple reports. ⁴⁷⁹ Surgeons at University College London Hospital said he could have lost his sight without the operation. ² That warning is the human stake that makes the technical detail legible: this was not presented as an experiment in computing, but as surgery to remove a tumour that threatened blindness. ²³⁴

The operation took place in May 2026 at the National Hospital for Neurology and Neurosurgery in London, part of University College London Hospitals NHS Foundation Trust. ¹⁰ The hospital is the specialist neurosurgical centre where the Trust concentrates complex brain cases, and the reports place both the surgery and the AI work there. ⁶⁷¹⁰ Reporting did not emerge until 26-27 August 2026, when BBC and Guardian pieces appeared, followed on 27 August by The Independent, Euronews, RTÉ, The National, TechSpot and Wionews, and on 28 August by the Times of India. ^2345781011 The three-month gap between surgery and publicity is not explained in the excerpts provided, but the timeline is consistent across the file.

What the AI is said to have done is narrow and specific, which is to its credit. The system analysed live camera footage during the procedure and helped the team identify nerves and blood vessels near the tumour. ⁶⁷¹⁰ One account says it flagged nerves and blood vessels for the surgical team as it worked around the tumour. ¹⁰ Another describes the real-time guidance as "colour-coded" to protect vital structures and save the patient's vision. ¹¹ In plain terms, the machine did not cut, decide, or operate; it watched the same field and labelled what should not be damaged.

Hibbert was described as the first patient in the world to have a brain tumour removed while medical teams used real-time AI during surgery. ²⁴⁷ The institutional source of the stronger formulation — the world's first successful AI-assisted operation of its kind — is UCLH itself. ⁶ That phrasing matters because the qualifier does most of the work: of its kind suggests live, in-theatre assistance for this type of neurosurgery, not any prior use of computers in planning, imaging, or retrospective analysis. The reports do not provide a published definition of the category, a registry search, or an independent adjudication of the firstness claim.

The outcome reported is binary and positive: the tumour was successfully removed. ²⁴⁷ The AI guidance is credited with saving his vision. ¹¹ What is missing is everything that usually surrounds a surgical innovation after the headline — no source in this set provides follow-up detail beyond the immediate reports, and no excerpts include recovery timelines, complication status, or verbatim comment from the surgical team or the patient. ²⁴⁷ The photograph record identifies neurosurgeon Hani Marcus alongside Hibbert in UCLH/PA images, but the excerpts do not supply a direct quotation from him or a detailed account of his role. ³⁴⁷⁸

That thinness should shape how the story is read. A single successful removal does not answer how a vision system behaves when blood obscures the lens, when anatomy varies, when image quality drops, or when a different surgeon in a different hospital must decide whether to trust a coloured overlay. The sources provided do not disclose a product name, a regulatory status, a trial protocol, or performance data from simulations and safety checks. They do not say how many hours of video trained the model, who funded it, or how control remained with the surgeons moment to moment, beyond the core description of live analysis and flagging of vessels and nerves. ⁶⁷¹⁰ Anyone filling those gaps from elsewhere is going beyond what this reporting file can support.

The temptation will be to treat a demonstration as a verdict. It is not. Surgery adopts tools that reduce uncertainty under pressure, and a tool that reliably labels hidden vessels and nerves in real time would be genuinely consequential if it proved faster, safer, or more complete across many cases. One labelled case proves that the team could run the system live without derailing the operation and that the patient left theatre with his tumour removed and his sight credited as saved. ²⁴⁷¹¹ The distance between those two statements — it ran live once, and it improves surgery — is the whole of the further research that has yet to be published.