How Smarter AI Could Reshape The Future Of Surgical Robots

Direct Source Verification: This story is aggregated from Forbes (forbes.com). Full reporting rights and copyright belong to the primary publisher.
The future of robotic surgery may depend not only on smarter AI but also on determining where each capability can be delivered most safely, reliably and economically.

Ming-Chien Chyu, PhD, PE, Founding President, Healthcare Engineering Alliance Society (HEALS), and professor at Texas Tech University.

gettyIn 2025, researchers reported SRT-H, a hierarchical AI system that autonomously performed steps of cholecystectomy on eight ex vivo gallbladders. While limited to a single procedure in an ex vivo setting, it suggests how substantial new capability can come from intelligence controlling an existing robotic platform.

That raises a broader design question: As AI takes on more functions, what should the surgeon, AI, robot and instrument each be responsible for?

I propose answering that question capability by capability, determining where each function belongs: with the surgeon, AI, robot or instrument.

Clinical judgment, oversight and critical decisions should remain centered on the surgeon. AI can increasingly assist with perception, planning, guidance and carefully bounded tasks. The robotic platform provides positioning, motion control and the mechanical stability, precision and safety functions that must remain physically enforced, while instruments provide physical interaction with tissue.

Where a capability belongs depends partly on whether it must work if higher-level software fails, whether it varies among procedures or patients, whether it can be reliably validated and monitored and how difficult it is to modify once physically built into a system.

Force sensing illustrates the trade-off. AI-based virtual sensing could potentially estimate some tool-tissue forces from information already available to the system, reducing reliance on dedicated sensors in appropriate applications. But estimated forces carry uncertainty and may suit some assistive functions better than safety-critical ones. Some capabilities may migrate toward AI; others remain fundamentally physical. Software cannot give an instrument a physical degree of freedom it lacks.

Mechanical limits, emergency stops and some redundant safeguards should remain independent.

AI can optimize within the safety envelope; it should not necessarily define it.

Nor does this imply one universal robot. Bone resection, laparoscopy and endoluminal procedures have different mechanical requirements. The principle is more realistically applied within compatible surgical classes: Decide where each capability belongs, then design the system accordingly.

Capability allocation and responsibility allocation are closely related. For every capability assigned to AI, developers, regulators and clinicians must also determine who has authority over that function and who remains accountable when it fails.

Telesurgery adds another dimension to the allocation question: where the AI itself should reside. Latency and network reliability remain important challenges in telesurgery. Surgical AI could reside with the remote expert, locally with the patient and robot, or be distributed between both locations. That choice could affect latency, autonomy, safety, responsibility and handoff.​

Smarter AI does not automatically reduce cost. Costs include instruments, service and training, while AI introduces development, validation, cybersecurity and life cycle costs. Moving versatility into software may lower some costs while creating others.

The real question is whether better capability allocation and smarter AI can improve efficiency, consistency and outcomes while reducing the total cost of delivering safe, high-quality robotic surgery.

Savings from reduced hardware complexity or operating-room requirements would need to outweigh additional software, regulatory and life cycle costs.

Research on autonomy in robotic surgery highlights limited access to high-quality surgical data, difficulties modeling deformable tissues and validating increasingly autonomous systems.

The challenge is not simply collecting more data, but building curated diversity in datasets with reduced errors while preserving clinically meaningful variations in anatomy, pathology, technique and unexpected events. Expert clinical annotation can help distinguish noise from meaningful variation, while multi-institutional data broaden representation. Simulation and synthetic data can supplement unusual scenarios but cannot yet fully reproduce biological complexity.

The FDA’s 2025 final guidance on predetermined change control plans (PCCPs) allows manufacturers to specify planned modifications to AI-enabled devices and how they will be developed and validated; changes outside an authorized plan may require additional regulatory review. The FDA’s 2026 draft guidance for robotically assisted surgical devices separately addresses controls, reliability, software, cybersecurity and other system requirements.

AI modularity therefore requires regulatory and validation boundaries as well as technical ones.

The key question is when autonomy should end and control should return to the surgeon.

The system should be able to detect and signal when it is operating outside its validated range, and then transition control according to a predefined handoff protocol. Determining when control should return to the surgeon remains a central challenge for autonomous robotic surgery.

In many proposed autonomous systems, an AI-controlled task could be required to stop when the AI’s own estimate of confidence falls below a validated threshold. But an AI system can be highly confident and still be wrong, and recognizing unfamiliar conditions remains challenging. Other potential handoff triggers include observed anatomy deviating sufficiently from expectation or an action approaching a predefined safety boundary. These are proposed triggers, not established clinical thresholds.

An equally important challenge is what happens after AI’s handoff. A surgeon taking control mid-task must rapidly understand what the system has done, why it stopped and what requires attention.

Autonomy is not defined only by what a system can do but also by how safely it gives control back when it reaches its limits.

The handoff function should therefore be treated as a safety-critical capability, not merely an AI feature.

A useful lesson may come from autonomous driving systems, where safe human takeover has become a central design challenge. In surgery, designing the handoff may ultimately prove as important as designing the autonomy itself.

The future of robotic surgery may depend not only on smarter AI but also on determining where each capability can be delivered most safely, reliably and economically.

β€’ What independent safeguard remains if the software fails?

β€’ Across what patient cohorts, procedures and conditions has it been validated?

β€’ What triggers the handoff, what does the surgeon see and what can the surgeon do when AI stops?

If developers cannot answer all three, that capability is not ready to move into AI.​

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?

Original Source
https://www.forbes.com/councils/forbestechcouncil/2026/11/19/how-smarter-ai-could-reshape-the-future-of-surgical-robots/
Visit Forbes β†—
SHARE STORY:
𝕏 f in

Related Coverage in Technology