Robot surgeons: how close are machines to operating alone?

Autonomous surgical robots are machines that carry out defined steps of an operation on their own, using their own movements. This is a field about machines performing parts of a procedure with limited human control. It is not about replacing surgeons. In the research described here, a surgeon still prepares the operation or approves the robot's plan, and none of the work involved patients. Nothing in this article describes routine clinical use.

The field matters for two reasons. Some surgical steps are precise and repetitive, so a machine that performed them consistently could, in principle, be useful. Each step towards autonomy also raises a harder question: who is responsible when a robot decides how to act inside a patient's body?

What has been demonstrated

The clearest early result came from a 2022 study in Science Robotics by a team that included Axel Krieger. A robot called STAR joined two pieces of pig small intestine using keyhole surgery. The authors report tests in phantom tissue, meaning artificial models, and in live pigs over a one-week survival period (Saeidi and colleagues, 2022). The live-animal part used five pigs. STAR operated on four, and a surgeon used manual laparoscopy on the fifth (NIBIB, 2022).

In that work, the surgeon first exposed the tissue. STAR then used images to propose a stitch plan based on the tissue's shape and thickness, and it stitched only after the surgeon approved the plan. If the tissue moved or deformed beyond a set threshold, the robot asked the surgeon whether to make a new plan. NIBIB does not give the threshold value. In the phantom tests, STAR performed the procedure five times, and four surgeons each performed it twice by manual laparoscopy and twice with a robot-assisted system. The synthetic intestine was moved back and forth to simulate breathing (NIBIB, 2022). The paper's abstract reports that, in the phantom tests, the autonomous system was more consistent and accurate than expert surgeons' manual technique and robot-assisted surgery, and that this was also seen in the live pigs (Saeidi and colleagues, 2022). NIBIB reports no observable difference in wound healing between the STAR and manual operations seven days after surgery (NIBIB, 2022).

A second study, published in July 2025, took a different approach. Kim and colleagues describe SRT-H, a framework in which a high-level planner works in language, producing task-level or corrective instructions, and a lower-level policy generates the robot's movements. The system was trained by imitation learning, which means learning from examples of the task (Kim and colleagues, 2025). It carried out a step of gallbladder removal (cholecystectomy) without human help on eight ex vivo gallbladders, meaning tissue outside the body, with a 100% success rate across those eight (Kim and colleagues, 2025; arXiv preprint). The abstract does not say what animal the tissue came from; Tech Xplore reports that the training videos came from surgery on pig cadavers. The abstract reports no live-animal results. Tech Xplore also reports that the robot took longer than a human surgeon, though its results were comparable to an expert surgeon's (Tech Xplore, 2025).

What is only reported

Some claims come from press coverage, institute summaries or the researchers themselves. These are useful context, but they are not measured results. NIBIB quotes Krieger saying that the results show STAR performing suturing more consistently and accurately than expert surgeons. It also quotes an NIBIB programme director describing the work as an important step towards autonomous robotic surgery in soft tissues (NIBIB, 2022). Krieger also compared the likely path to driver-assistance features that lead towards self-driving cars, and acknowledged that many people may be uneasy about a machine doing this work.

Tech Xplore says the SRT-H robot "operated for the first time on a lifelike patient" (Tech Xplore, 2025). The study's own abstract describes ex vivo tissue, not a person (Kim and colleagues, 2025), so this phrase should not be read as human use.

What is speculative

Some future uses have been suggested but not tested. Krieger has said autonomous robots could help make surgical care more widely available (NIBIB, 2022), and the STAR paper's abstract says such robots have the potential to improve consistency, patient outcomes and access to a standard technique (Saeidi and colleagues, 2022). The SRT-H authors describe their work as a milestone towards clinical use (Kim and colleagues, 2025), and Tech Xplore reports that the team wants to extend it to a complete autonomous operation (Tech Xplore, 2025). These describe directions for research, not outcomes.

Limits and open problems

No studies in people. None of the sources reviewed describes an autonomous surgical robot used in patients. The published autonomous work uses phantoms, live pigs and gallbladders removed from the body.

Small samples. The results come from small research teams and small numbers of animals or tissue samples. The sources reviewed do not describe independent groups repeating the work in other hospitals or with other surgeons.

Narrow tasks. SRT-H covered one step of one operation. A complete autonomous operation has not been shown.

Unexpected changes. The STAR system relied on the surgeon to expose the tissue and to decide on a new plan when the tissue moved. That suggests that handling unexpected changes alone is still an open problem. The gallbladder work was also slower than a human surgeon.

Regulation. A 2024 systematic review of 49 surgical robots cleared by the US Food and Drug Administration between 2015 and 2023 found that 86% were at level 1 of an autonomy scale, and 6% at level 3 (Lee and colleagues, 2024). In September 2026 the FDA issued draft guidance, "Robotically-Assisted Surgical Devices - Premarket Submissions", which is not yet in force (FDA, 2026). According to a summary by the law firm Alston & Bird, the draft focuses on local teleoperation, where a surgeon at a console in the same room controls the robot, and excludes autonomous robots that independently perform significant parts of a procedure (Alston & Bird, 2026). That is a secondary reading, not the regulator's own summary. The sources reviewed do not describe an approval route for fully autonomous surgical robots.

Liability and ethics. The sources reviewed do not settle who is responsible when a robot makes a decision during an operation. A 2017 article in Science Robotics argues that regulatory, ethical and legal barriers need careful consideration of different levels of autonomy and the context in which a robot is used, and proposes six levels of autonomy for medical robots (Yang and colleagues, 2017). PubMed indexes it as an editorial. Only its abstract was checked here, not the full text.

What to watch next

• A peer-reviewed study in which a robot completes a whole operation on its own in living animals.
• Independent research groups repeating autonomous surgical steps, in living animals or on other procedures.
• Any study in people, with defined human supervision and clear stopping rules.
• The FDA's comment period on the draft guidance, which closes on 24 November 2026, and a public workshop on 2 and 3 December 2026 that Alston & Bird says will cover robots with autonomous or remote teleoperation capabilities.
• Final FDA guidance, or a separate pathway, that explicitly covers fully autonomous surgical robots.
• Published comparisons of time, complications and tissue outcomes between autonomous and surgeon-led steps.

This article is general information, not medical advice.