Surgeons just performed the world's first live AI-assisted brain surgery

There are no road signs inside the human skull, and one millimeter in the wrong direction can mean the difference between saving a patient and blinding them.

Now, a 48-year-old father from Bedfordshire named Rhys Hibbert has become the world's first patient to undergo brain surgery guided by real-time artificial intelligence. Surgeons at University College London Hospital used the system to remove an 11mm pituitary tumor that had been pressing against his optic nerves and threatening his vision.

Eight weeks after the operation, Hibbert says his peripheral vision is back, his dizzy spells are gone, and he felt like he woke up with 360-degree sight.

A second pair of eyes behind the nose

The pituitary gland sits at the base of the skull, surrounded by critical carotid arteries and optic nerves. Surgeons reach it by feeding an endoscope up through the patient's nose, navigating behind layers of bone and membrane where every person's anatomy looks slightly different.

Historically, the stakes have been steep. Surgeries like this carry a 25% to 50% chance of leaving part of the tumor behind, and a 0.5% to 2% risk of nicking a major blood vessel.

Most UK surgeons only perform around 10 to 20 of these procedures a year, largely relying on pre-op scans to build a mental map before cutting.

The new AI, developed by UCL researchers with backing from Google and the UK National Institute for Health and Care Research, acted as a live co-pilot. Running on a second monitor, it analyzed the endoscope's video feed in real time, tracked instruments, and mapped out where hidden vessels and nerves were hiding beneath tissue.

Lead surgeon Prof Hani Marcus likened it to smartphone facial recognition, but tuned for human anatomy. Trained on hundreds of meticulously annotated surgical videos, Marcus noted the AI has seen more operations than most surgeons perform in an entire career.

Crucially, the system offers visual guidance, not physical intervention. Surgeons can glance at the overlay for advice or ignore it completely.

Why it matters

For patients facing high-stakes skull-base surgery, a real-time safety net could dramatically cut down on catastrophic complications and reduce the odds of needing follow-up operations.

We think UCL's approach is brilliant because of what it does not try to do. Instead of building multi-million-dollar surgical robots that take up entire operating theaters, the team built a purely software-driven overlay that plugs into existing endoscopes. That drastically lowers the barrier to adoption for hospitals worldwide.

While previous medical AI has mostly been confined to offline radiology reports and static pre-op planning, moving vision models into live, latency-sensitive surgical feeds marks a massive shift in how clinical AI gets deployed.

That said, we see real risks lurking here. The AI is still inferring hidden anatomy based on probabilities from past datasets; if it encounters an extreme anatomical anomaly, a false green zone could prove disastrous. There is also the quiet threat of cognitive offloading: if the next generation of surgeons grows up relying on software highlight reels, preserving their own spatial instincts in edge cases might become a serious challenge.