What the FDA Wants Surgical Robots to Prove
Surgical robots are usually judged by what they do in the operating room.
The FDA judges them by what the manufacturer can prove before they get there.
On September 25, 2026, the agency released draft guidance on premarket submissions for robotically-assisted surgical devices, which it abbreviates as RASDs.
The document covers testing, clinical evidence, software, cybersecurity, training, and labeling.
It is a nonbinding draft that is not for implementation, and the FDA is accepting comments through November 24, 2026, under docket FDA-2026-N-9505.
What the FDA is describing
The guidance covers robots in which a surgeon sits at a console and controls instruments held by arms beside the patient.
The FDA treats the console, the arms, and the operating room equipment as one connected system.
Problems can come from the connections between those parts, such as delays or unexpected arm movement.
Testing is therefore expected for each part and for the whole system working together.
What manufacturers should prove
The delay between the surgeon's hand movement, the instrument, and the image on screen should stay within safe limits, even under heavy load.
Safety stops should be measured by how far the instrument keeps moving after the stop is triggered.
Emergency undocking should be timed, so clinicians know how quickly they can reach the patient.
Cybersecurity testing should include a simulation of the robot connected to a compromised hospital network.
Performance should be shown on at least three systems, which proves the results are repeatable and not a one-time success.
What clinical evidence looks like
A robot that is new to the U.S. market will often need clinical data.
An existing robot may need it for a new use, a new patient group, a new technology, or a claim such as superiority.
The FDA suggests measures such as length of stay, complications during and after surgery, transfusions, conversion to another surgical approach, readmissions, reoperations, mortality, and operating time.
Comparisons with other robotic surgery are preferred over comparisons with open or laparoscopic surgery.
What it means for new companies
A working prototype is not enough, because the testing is meant to reflect systems built through production.
A validated training program for the whole surgical team is expected before clinical use.
A company seeking several procedures can use data from a complex operation to support simpler, related ones in the same specialty.
Where autonomy fits
The guidance defines autonomy as a robot performing tasks and making decisions from its own sensing without user intervention.
It excludes robots that perform significant parts of an operation independent of a qualified practitioner, and it excludes remotely operated systems.
AI features inside a covered robot, such as identifying critical structures, are in scope and call for an early conversation with the FDA.
The FDA has scheduled a public workshop for December 2 and 3, 2026 on robots with autonomous or remote operation capabilities.
The guidance is not the FDA's rulebook for autonomous surgery, because it covers supervised systems and leaves more autonomous ones to further regulatory work.
Readers tracking how vendors describe their systems against the six-level ladder can see where nine robots land in The Surgical Autonomy Registry.
What hospitals should ask
Training expectations cover bedside assistants and scrub nurses as well as the surgeon.
Labeling should list each compatible accessory by manufacturer, name, and model.
For hospitals evaluating a platform, the guidance becomes a short list of questions for vendors.
How much delay does the system have under worst-case conditions?
How long does emergency undocking take?
What cybersecurity testing has actually been performed?
The bottom line
The draft does not say which robot to buy, and it does not guarantee authorization.
It makes the evidence expected around surgical robots explicit.
The question that matters is less whether a robot can perform the procedure and more whether its maker can prove how it behaves when things go wrong.
The guidance is a draft and may change before it is finalized.
