The role and risks of AI in the future of surgical robotics

Abstract

This debate examines a pivotal question facing surgical robotics: whether the rise of universal, open-weight foundation models, capable of piloting diverse robotic systems will democratize access to advanced surgery or instead amplify existing ethical and social risks. 

Recent work points to an emerging ‘algorithmic surgeon’ paradigm in which a single, hardware-agnostic ‘brain’ can generalize across tasks and platforms. Proponents argue that open-source surgical AI (e.g. Gr00t-H) could dismantle the commercial moats built by proprietary hardware ecosystems, enabling competition, lowering costs, and accelerating clinical innovation. If the intelligence layer becomes modular and widely available, hospitals and researchers may iterate faster, share best practices, and expand training opportunities beyond a handful of elite centers.

Yet the same shift toward semi-autonomous operation raises urgent concerns about consent, equity, and accountability. Informed consent is challenged when model reasoning is opaque: if neither patient nor surgeon can reliably explain why the system chose a particular action, consent risks becoming a procedural formality rather than a meaningful choice. Equity concerns are similarly two-edged. While open models can reduce software barriers, the capital and operational costs of robotic infrastructure, data collection, cybersecurity, and specialized staff may keep advanced care concentrated in wealthy systems, potentially deepening disparities across regions and populations. Dataset pooling, even across many institutions, can still encode bias if coverage is uneven, clinical practices are homogenized, or underrepresented groups remain sparsely sampled.

The debate also confronts a looming liability vacuum as autonomy increases. AI is not a moral agent, yet errors will occur, and when they do, should responsibility fall on the surgeon as the last human in the loop, on the hospital as deployer, on developers who trained and tuned the model, or on hardware manufacturers who set operational constraints? 

Finally, the panel explores whether automation erodes the therapeutic relationship by distancing patients from human judgment, or whether it can instead free surgeons to invest more time in communication, empathy, and shared decision-making. By weighing commercial dynamics, regulatory realities, and core medical principles, this discussion aims to clarify what a responsible, and socially legitimate path to autonomous robotic surgery could look like over the coming years (or decade?).

Questions for the panel

Equality

1. Robotic surgeries currently disproportionately benefit wealthy, white populations due to immense capital infrastructure costs. Will open-source medical foundational models democratize global access to advanced surgery, or will the steep infrastructure and training requirements simply entrench existing healthcare disparities?

2. AI models are highly susceptible to algorithmic bias when trained on non-representative data, risking worse outcomes for marginalized groups. Does pooling data into open datasets (e.g., Open‑H across 49+ institutions) meaningfully reduce bias, or does it instead create a homogenized model that misses individual patient nuance and encodes hardware/platform-specific bias from the robots, instruments, sensors, and control systems used to collect the data?

Trust

3. Clinical AI tools used in surgery (decision-support, perception, guidance, and safety systems) can still function as “black boxes,” making it difficult for patients, and sometimes clinicians, to understand how recommendations are produced. As surgical robotics progress toward greater automation, if surgeons cannot explain what the system is doing, its uncertainty, and its failure modes, does informed consent become more procedural than truly informed? 

Liability

4. As surgical AI becomes more capable, adverse events will trigger disputes about responsibility. Since AI isn’t a self-aware “moral agent,” what is the appropriate liability model, surgeon as operator, hospital as deployer, developer as manufacturer, or OEM as system integrator, and what evidence (audit logs, validation scope, human override expectations) should determine fault?

5. Current legal frameworks are ill-equipped to handle AI-involved surgical errors, and 86% of FDA-cleared robots remain at LASR Level 1. Given the strict mandates of the EU AI Act and the FDA’s cautious approach to foundation models, is clinical deployment of an autonomous VLA a realistic goal for the 2030s?

Business Models

6. If open-weight medical foundational models can eventually pilot any robotic arm, does surgical hardware collapse into a low-margin commodity? Or will proprietary incumbents successfully rebuild their commercial moats around specialized instrument ecosystems and tightly controlled clinical training networks?

7. AI models are built on the “digital distillation” of surgical intuition and patient data, yet the resulting commercial value often flows almost exclusively to device manufacturers. How should the intellectual property of a surgeon’s skill be protected in an era of mass data harvesting, and what models—be they royalties, data-unions, or shared IP—must exist to ensure that clinicians are not economically displaced by the very models trained on their own expertise?

8. Closing: is a universal, hardware-agnostic surgical AI the inevitable future of medicine, or an ethical and technological abstraction doomed to fail in the unforgiving reality of the operating room?