> ## Content Index
> Fetch the complete content index at: https://nexi.fund/llms.txt
> Use this file to discover other available public pages before exploring further.

# AI Copilots for Robotic Surgery Reach a Clinical Milestone
- URL: https://nexi.fund/ai-surgical-copilot-2026/
- Published: 2026-07-08T10:30:37.000Z
- Updated: 2026-07-08T10:30:37.000Z
- Description: An AI copilot for robotic surgery completed its first live clinical demonstration in Ghent. The surgical robotics market reached $18.3B, with AI-enabled systems capturing 35% of new installations. At least four research groups now field working prototypes.
- Author: Nexi.fund Labs
- Tags: AI & Infrastructure, Biotech & Health, #mode-6, #hook-question, #track-F

What happens when the surgeon's most experienced assistant is not a person, but a vision-language model that has studied 23,000 procedures before walking into the operating room?

🎯

An AI copilot for robotic surgery, a vision-language-action model that recognizes anatomy, predicts next steps, and guides instruments in real time, completed its first live clinical demonstration in Ghent, February 2026.  
  
The surgical robotics market reached $18.3 billion globally, with AI-enabled systems capturing 35% of new installations.  
  
At least four independent research groups and three medical-device companies now field working AI copilot prototypes. A pace that challenges FDA's current framework for continuously learning systems. 

For two decades, surgical robotics meant one thing: the da Vinci system, a mechanical extension of the surgeon's hands. The robot held the tools. The surgeon made every decision.

That model is breaking apart. In 2025 and 2026, the industry crossed into a different territory, one where the machine does not just hold instruments but watches, interprets, and suggests the next cut.

$18.3B Surgical robotics market, 2026 ↑ 35% AI-enabled share 

#### AI-enabled surgical systems capture market share

RoboCloud Hub estimates that AI-capable platforms now account for more than a third of new surgical robot installations globally. Intuitive Surgical's da Vinci still leads in absolute numbers, but Medtronic's Hugo and CMR Surgical's Versius, both designed with AI-assisted features from the ground up, are closing the gap at a faster rate than earlier forecasts predicted. *RoboCloud Hub, Dec 2025*

## The first live AI copilot in the OR

In February 2026, at the Surgical AI & Telesurgery Days in Ghent, Orsi Academy demonstrated what it calls a world first: an AI copilot that analyzes surgical video in real time, recognizes anatomical structures, identifies the current phase of the procedure, and surfaces contextual guidance without requiring the surgeon to shift focus.

The system was developed by Orsi Innotech, a spinout led by urologist-engineer Pieter De Backer. It runs on NVIDIA's medical supercomputing infrastructure, a setup purpose-built for the latency constraints of live surgery.

"The innovation is not a future scenario, but a concrete solution that can be put into practice," said Prof. Dr. Alex Mottrie, CEO of Orsi Academy.

De Backer's team had already demonstrated the underlying approach in 2023, deploying four AI models simultaneously during a robot-assisted kidney tumor removal, combining organ recognition, phase detection, anonymization, and augmented reality, without clamping renal blood supply. The 2026 copilot integrates those capabilities into a single, continuously operating system.

## What an AI copilot actually sees

The technical shift is worth unpacking. Earlier surgical AI models operated as isolated classifiers: one model identifies a needle, another segments an organ, a third flags bleeding. The copilot model connects them into a reasoning loop.

The RARP Copilot, a vision-language model for robot-assisted radical prostatectomy (RARP) published in npj Digital Surgery in April 2026, demonstrates this architecture. It takes live video from the endoscope, translates visual features into structured scene descriptions, and generates action proposals. The system not only identifies what it sees but predicts what should happen next.

In endoscopic surgery, a team at npj Digital Medicine showed in June 2026 that vision-language-action (VLA) models can integrate multimodal cues, camera feed, instrument telemetry, patient vitals, to infer hidden tissue dynamics. "Properly implemented, reasoning-driven autonomy can transform AI copilot robots from reactive executors into cognitive collaborators," the authors wrote.

The model does not replace the surgeon. It compresses the cognitive loop: instead of the surgeon watching, interpreting, deciding, and acting, the copilot watches, interprets, and presents options. The surgeon decides and acts, with less lag and less mental load.

📊

**Key signals to track**  
  
FDA clearance of a continuously learning surgical AI. No precedent exists for a system that updates its model post-deployment.  
Clinical trial data comparing AI-copilot-assisted outcomes against standard robotic surgery. The first such trial is expected to read out in late 2026 or early 2027.  
A surgical Vision-Language-Action (VLA) model trained on cross-institutional data, analogous to what RT-2 did for general robotics  
Any liability ruling or regulatory guidance on who bears responsibility when an AI copilot's recommendation leads to patient harm 

## From Intuitive's monopoly to a fragmented market

For years, surgical robotics was simple: da Vinci owned it. Intuitive Surgical's platform, first cleared by the FDA in 2000, accumulated the largest installed base, the most trained surgeons, and the deepest moat in procedural data.

That moat is thinning. Medtronic's Hugo robot received CE mark for general surgery in 2024 and has since expanded indications. CMR Surgical's Versius passed 10,000 clinical procedures in 2025 and launched an AI skill-assessment module. Shanghai MicroPort's Toumai robot achieved 100 commercial orders in early 2026 and, in a December 2025 milestone, performed the first fully autonomous abdominal surgery on a live animal, completing 88% of surgical steps on the first attempt without human intervention, using a proprietary multimodal AI called Neuron trained on 23,000 recorded procedures.

The competitive dynamic has shifted from "who builds the best robotic arm" to "who builds the best AI layer on top of it."

## Autonomy in stages: where the bar actually is

Fully autonomous surgery on humans has not happened yet. The Shanghai MicroPort pig trial was a proof of concept, not a clinical deployment. The STAR system from Johns Hopkins, which demonstrated autonomous laparoscopic suturing with 50% faster completion and 40% fewer errors than expert surgeons, operates under research protocols with a surgeon in the loop who can interrupt in under 100 milliseconds.

The practical milestone for 2026 is not Level 5 autonomy. It is the AI copilot that stays in an advisory role but is reliable enough that surgeons begin to trust its suggestions and, over time, to act on them without independent verification of every recommendation.

```

AI SURGICAL COPILOT — MATURITY LADDER
─────────────────────────────────────────────────────────────
  TRL 1–3        TRL 4–6        TRL 7–8        TRL 9
  ✓        ──── ✓         ──── ◉ NOW     ──── 
  Research       Pilot          Scale-up       Market
─────────────────────────────────────────────────────────────
[████████░░] 80%  ·  Copilots in clinical demo / research use; first commercial OR deployment expected 2027
Source: Yazamaz, npj Digital Surgery, 2026

```

## The compute pipeline behind the copilot

Running a vision-language-action model in real time during surgery imposes infrastructure requirements that few hospitals currently meet. The Orsi Academy copilot runs on an NVIDIA supercomputer, purpose-built for the latency constraints of an operating room where a 500-millisecond delay can be clinically significant.

NVIDIA's Clara platform, originally developed for medical imaging, has been adapted to support surgical AI workloads. The company now positions its DGX systems as the compute backbone for what it calls "digital surgery": real-time inference, video processing, and model updates running on the same hardware stack.

The data pipeline is equally demanding. Shanghai MicroPort's Neuron model trained on 23,000 recorded procedures. The STAR system at Johns Hopkins drew on thousands of hours of laparoscopic video. No single hospital generates enough surgical data to train a production-grade copilot, which means effective models will require cross-institutional data sharing, with all the privacy, consent, and competitive barriers that entails.

Orsi Academy's approach addresses this indirectly: the copilot runs inference locally on the NVIDIA system but the model improves through centrally aggregated, de-identified training data from multiple partner hospitals. That architecture, distributed inference combined with centralized training, mirrors the pattern already established in autonomous driving and general-purpose robotics.

The harder question is whether the surgical community can replicate the data-sharing willingness that made autonomous driving possible. Hospitals compete for patients. Surgeons compete for reputation. Device manufacturers compete for installed base. Each has reasons to keep procedural data private.

## FDA on continuously learning systems

The FDA issued draft guidance for AI-enabled medical device software in January 2025, foregrounding transparency, model inputs and outputs, performance bias disclosure, and the concept of a "predetermined change control plan." The guidance addresses a problem unique to surgical AI: how do you regulate a system that improves with every procedure?

A copilot that trains on 50,000 surgeries is safer than one trained on 5,000\. But the path from 5,000 to 50,000 passes through patients whose data the model uses to update itself, without a new premarket submission each time.

The FDA is aware of the tension. The draft guidance has not yet been finalized. No continuously learning surgical AI has received clearance. RecovryAI, a post-surgical monitoring platform rather than an intraoperative copilot, received FDA Breakthrough Device Designation in March 2026, a signal that the agency is engaging with the category, but not yet approving it.

## The open question

The AI copilot for surgery is not science fiction. It was demonstrated live. It has been published in peer-reviewed journals. Multiple companies are building competing versions. The technology works in controlled settings.

What has not been proven, and will not be proven by laboratory results alone, is whether it works better than a skilled surgeon with a standard robotic platform, across diverse patient populations, in hospitals with varying data infrastructure, under the legal and regulatory frameworks that currently exist.

A January 2025 Science Robotics review framed the challenge precisely: "AI and robotics are enabling surgical autonomy, with the potential to improve outcomes and expand access to care. The most balanced reading is that AI has matured today to augment rather than replace the surgeon's judgment."

That is where the field sits in mid-2026\. The copilot has joined the crew. It has not taken the controls.

[ How can reasoning capability empower the AI copilot robot in endoscopic surgery Vision-language-action model architecture for surgical AI copilots — covers the reasoning loop that transforms reactive tool-holders into cognitive collaborators. npj Digital Medicine (Nature) ](https://www.nature.com/articles/s41746-026-02827-8?ref=nexi.fund) 

Technical deep-dive on VLA model logic for endoscopic AI copilots

[ Artificial intelligence for the future of digital surgery Editorial framework for the emerging AI copilot ecosystem, covering the three levels of assistance: pre-operative planning, intra-operative guidance, and autonomous sub-task execution. npj Digital Surgery (Nature) ](https://www.nature.com/articles/s44484-026-00008-4?ref=nexi.fund) 

Comprehensive editorial overview of the surgical AI copilot landscape

[ Surgical RARP copilot: a vision language model for robot-assisted radical prostatectomy Published implementation of a working VLM copilot for prostate surgery — real-time scene understanding and action proposal generation from endoscopic video. npj Digital Surgery (Nature) ](https://www.nature.com/articles/s44484-025-00003-1?ref=nexi.fund) 

Concrete architecture and results from a deployed surgical copilot prototype