When Optimus Operates

In January 2026, Elon Musk told podcast host Peter Diamandis that Tesla's Optimus humanoid robot would become a better surgeon than the world's best human surgeons within three years, adding: "There will probably be more Optimus robots that are great surgeons than there are all surgeons on Earth"[1].

The claim is not about a specialised surgical instrument. Optimus is a general-purpose, bipedal humanoid robot originally developed for factory work and household tasks, not a device engineered, tested or classified for clinical use. Musk's timeline would have that same general-purpose robot walk into an operating room and outperform every human surgeon on the planet, at global scale, inside three years.

The most advanced autonomous surgery achieved to date looks very different from that vision. In July 2025, researchers at Johns Hopkins University and Stanford University reported that a fixed, purpose-built surgical robot completed the clipping-and-cutting phase of a gallbladder removal, a minutes-long sequence of seventeen tasks including identifying ducts and arteries, placing clips and severing tissue, entirely on its own and with a 100% success rate across eight ex vivo gallbladders[2]. The robot adapted to non-uniform anatomy and recovered from unexpected changes introduced mid-procedure, a genuine technical milestone. But it was still one defined phase of one procedure, performed on non-living tissue by a stationary system engineered for nothing but surgical manipulation, a considerable distance from a general-purpose humanoid performing open-ended surgery on a living patient.

Bioethicist Arthur Caplan of NYU Grossman School of Medicine called Musk's three-year claim "not credible," citing the difficulty of programming for human anatomical variability and the years of outcome data needed to demonstrate equivalence with human surgeons[1]. A systematic review of every surgical robot cleared by the U.S. Food and Drug Administration between 2015 and 2023 supports his scepticism: 86% of cleared systems still operate at the most basic level of autonomy, where a surgeon initiates and directs every motion, and none has reached full autonomy[3].

The distance between a rhetorical target and a demonstrated capability is itself a risk. When a company's public roadmap for a general-purpose robot includes performing surgery on human beings, the question is no longer only whether the technology will eventually work, but what happens in the years between the announcement and the evidence.

When the Demonstration Does Not Match the Claim

At a Tesla "Autonomy Visualized" event in Miami in December 2025, an Optimus unit handing out water bottles suddenly raised both hands to its head in a motion widely likened to someone removing a virtual-reality headset, lost its balance, and fell backward in front of attendees. Robotics engineers and industry observers concluded the robot had been under teleoperation and that its remote operator had removed their headset mid-task[4]. The episode was not an isolated one: at Tesla's "We, Robot" event in October 2024, Optimus units seen dancing, mixing drinks and conversing with attendees were also confirmed to be under human teleoperation, with Morgan Stanley analyst Adam Jonas writing that the robots "relied on tele-ops (human intervention)"[5].

None of this means Optimus cannot improve quickly. It does mean that, as of the claim's own three-year clock starting to run, the robot could not reliably serve drinks at a controlled promotional event without a human operator, let alone operate on a patient.

Writing in direct response to Musk's prediction, head and neck surgeon Narayana Subramanian argued that "the fundamental challenge in surgery is not precision but adaptability," and that autonomous systems struggle precisely where surgery is hardest: unpredictable anatomy, prior treatment history and intraoperative judgment calls that fall outside a model's training data[6]. No fully autonomous surgical system has reached routine clinical deployment, and FDA-cleared surgical robots remain far from Level 4 or Level 5 autonomy[3]. Subramanian also noted that standard metrics for evaluating such systems are lacking, and that the distribution of liability between manufacturing defects, medical negligence and software failure remains unresolved even in the most specialised, purpose-built surgical devices[6].

A general-purpose humanoid robot compounds every one of those problems. It must additionally solve safe bipedal movement around a sterile field, dexterous manipulation with hands not designed for surgical instruments, and reliable operation without the remote human assistance its own public demonstrations have depended on.

A Robot With No Medical Classification

Existing surgical robots, however limited their autonomy, are at least regulated as medical devices for a defined intended use. In the United States, FDA device classification depends on intended use, indications for use and risk; most FDA-cleared surgical robots reviewed from 2015 to 2023 entered the market through the 510(k) pathway, with a growing number introduced through the De Novo pathway[3]. Optimus has no such classification of any kind, because it was not built, submitted or cleared as a surgical device; it is being developed and marketed as a general-purpose robot for factories and homes.

A 2017 framework proposed by a group of leading medical robotics researchers warned that this problem grows sharper, not simpler, as autonomy increases. At the highest levels of autonomy, they argued, a medical robot is no longer simply a device the FDA can clear; it is "practicing medicine," an activity the FDA does not regulate at all, since medical practice is governed by state licensing boards and medical societies built around human practitioners[7]. A general-purpose humanoid robot performing surgery sits in neither system cleanly: not a licensed practitioner, and not a device cleared for the clinical task it would be performing.

The liability question is equally unresolved. Medical professional liability insurance is built around a physician who can be named as the insured and whose conduct can be measured against a standard of care. Traditional policies were not written around a non-human actor making autonomous clinical decisions, which leaves open questions about who is insured, whose act triggers coverage, and how the timing of the harmful event would even be analysed[8]. Those questions are already difficult for a purpose-built surgical device; they are harder still for a general-purpose robot with no medical credential, no clinical trial record and no device clearance for the procedure it performed.

Before the Robot Can Operate, the Rules Must Catch Up

A future in which a humanoid robot extends expert-level surgical care to places that lack it today would be a genuine achievement. Getting there safely requires the technology, the evidence and the regulatory and liability framework to advance together, rather than letting a public timeline for the first outrun the other two.

Before any general-purpose humanoid robot performs surgery on a human being, there are questions that need answers, not announcements: what clinical trial evidence demonstrates the robot's performance against human surgeons across real anatomical variation, what regulatory pathway and device classification the robot must pass before it is allowed near a patient, and which party, manufacturer, hospital, software provider or insurer, bears responsibility when it fails. Until those questions are answered, the gap between what a humanoid robot has been shown to do and what it has been promised to do is where the risk lives.