FDA published draft guidance on premarket submissions for robotically-assisted surgical devices in the Federal Register on September 25. Comments close November 24 under docket FDA-2026-N-9505.
The scope is narrow on purpose. It covers teleoperated, software-controlled systems built from an operator console, a bedside subassembly carrying the working arms and instruments, and a subassembly for OR staff. It leaves out preoperative planning software, stereotaxic systems, interventional device placement systems, remotely teleoperated systems, and robots that act without a practitioner at the controls.
The testing asks are specific. Total system latency between hand input, instrument motion and the displayed image has to be measured under worst-case conditions. FDA wants that done on at least three devices, against a limit the sponsor justifies clinically. Accuracy, repeatability and resolution also take three devices. Grip and cut reliability should reach 95 percent reliability at 95 percent confidence. Instrument testing runs through the labeled reprocessing cycles out to the end of use life. Visualization performance follows IEC 60601-2-18. The three-device minimum is the line that moves a build schedule, not just a test plan.
FDA expects Enhanced software documentation for most of these systems, plus a cybersecurity assessment under section 524B that reaches back-end infrastructure and cloud services, with penetration testing that assumes the hospital network is already compromised. For De Novo requests the draft floats an umbrella and covered procedure approach, where data from a higher-complexity procedure can support authorization for the lower-complexity ones underneath it. Labeling has to name every compatible accessory and third-party device by manufacturer, version and model number. Separately, FDA set a December 2 to 3 public workshop at White Oak on benefit and risk for the autonomous and remote teleoperated systems this draft excludes. In-person registration closes November 23.