Surgical Artificial Intelligence Innovation Trends and Regulatory Insights

Objective: To characterise the landscape of surgical artificial intelligence (AI) devices approved by the US Food and Drug Administration (FDA). Background: AI has the potential to improve surgical pathways, offering capability across screening, diagnosis, decision support, and intraoperative care. Despite this, real-world adoption remains limited. Regulatory approval serves as a key step in clinical translation but is influenced by challenges in evidence generation, safety, bias, transparency, and implementation. Consequently, the characteristics, evidence base, and functional roles of FDA-cleared surgical AI devices remain incompletely defined. Methods: FDA-cleared AI/machine learning medical devices were systematically screened for surgical relevance up to July 2025. Eligible devices were extracted for clinical application, underlying AI technology, evidence quality, and regulatory pathway. Devices were categorised using an adapted descriptive, diagnostic, predictive, prescriptive framework, and predicate networks were evaluated for risk of predicate creep. Analyses were performed in R. Results: A total of 314 devices (25.1%) had surgical applications. Most were imaging-based devices and mapped to general surgery (24.8%), orthopaedic surgery (18.2%), neurosurgery (17.2%), and cardiac surgery (10.8%). The majority of devices were descriptive (50.6%) or diagnostic (40.4%), with relatively few predictive (1.3%) or prescriptive (7.6%) applications. Clinical validation was predominantly based on retrospective or bench studies, with only 5.1% supported by higher-quality prospective evidence. The premarket notification 510(k) pathway accounted for 95.2% of approvals. Among these, 37.5% demonstrated a high risk of predicate creep, reflecting substantial differences in intended use or technological characteristics compared with predicate devices. Conclusion: FDA-approved surgical AI devices are currently dominated by imaging-based diagnostic tools. The widespread reliance on the 510(k) pathway and the observed prevalence of predicate creep highlight important considerations for regulatory evaluation. As AI applications expand toward predictive and interventional roles, improved evidence transparency and robust assessment of device equivalence will be essential to support safe clinical adoption.