Surgical data extractivism: the false promise of institutional custodianship

As artificial intelligence (AI) rapidly evolves, maturing AI methods have the potential to optimize perioperative surgical care, improve patient outcomes and facilitate surgical education (1). Recent studies have also demonstrated the feasibility of integrating AI into the clinical practice of hepatopancreatobiliary surgery. For instance, AI models enable the early diagnosis and perioperative risk stratification of hepatocellular carcinoma (HCC). Furthermore, AI algorithms have been utilized to identify anatomical structures in the surgical field, facilitating real-time intraoperative decision support. They also show excellent performance in guiding treatment allocation for recurrent HCC after surgery (2-4). However, the rapid emergence of these surgical AI models poses a critical challenge regarding data governance, raising the fundamental question of who should ultimately govern the sensitive patient information required for these technologies.

Gorijavolu and Sakran’s recent viewpoint “Who Owns the Operating Room? Surgical Data Governance at a Crossroads” published in JAMA Surgery proposes institutional custodianship as the ethical foundation for surgical data governance (5). Their framework assumes that hospitals and health systems inherently possess the structural incentives to serve as neutral fiduciary trustees of patients’ data and privacy. However, as AI transitions from theoretical computational modeling to clinical reality, emerging analyses of health system commercialization demonstrate a concerning trend that fiscal pressures increasingly position medical institutions as data brokers rather than protective stewards. This issue is particularly urgent within surgery.

Indeed, this dynamic is largely driven by the fact that the development of sophisticated AI models fundamentally relies on the integration of massive, high-fidelity clinical datasets (6). These datasets, comprising electronic health records, multiphase preoperative imaging, digitized histopathological slides, and continuous intraoperative video, are increasingly transformed into liquid assets. They are frequently extracted and exchanged for vague research partnerships that ultimately train proprietary algorithms locked behind foreign paywalls. When the designated custodian of the data is concurrently acting as its seller, fiduciary duty inevitably dissolves into a commercial transaction. This environment fundamentally transforms the operating room from a sanctuary of healing into a resource frontier, where the most vulnerable moments of patient care are systematically harvested for commercial gain.

The Global South bears a disproportionate hepatopancreatobiliary disease burden, mostly driven by the endemic prevalence of infectious etiologies (7). High-volume surgical centers in these regions serve as the primary engines generating the raw data required to train robust AI models. Consequently, these high-fidelity datasets flow unidirectionally from the Global South into Western AI development pipelines, exemplifying surgical data extractivism, or data colonialism (8). The resulting algorithms, despite being trained on the diverse anatomical and pathological presentations of developing populations, are priced exclusively for premium Western markets. They eventually return to the originating institutions accompanied by prohibitive licensing fees that severely strain already fragile public health budgets. Institutional custodianship merely sanitizes and legitimizes this neocolonial economy, wherein developing regions mine raw biological data solely for the Global North’s technological profit.

Furthermore, achieving true de-identification of surgical data, often emphasized by institutional review boards, is highly challenging in the era of advanced computer vision (9). The core of high-fidelity clinical datasets, intraoperative video, constitutes a distinct form of biometric property that is fundamentally resistant to true anonymization. Individual operative technique, instrument handling patterns, and patient-specific anatomy serve as persistent identifiers, making passive transparency insufficient. Governance must enforce rigorous medical algorithmic auditing and verifiable data provenance through technologies such as cryptographic ledgers to prevent retrospective manipulation (10). Furthermore, independent multidisciplinary committees must practice continuous algorithmic stewardship by utilizing post-deployment surveillance to monitor for algorithmic drift and ensure AI models maintain clinical equity across diverse populations.

These factors explain why institutional custodianship is inherently flawed. However, simply transferring data sovereignty to surgical societies, a model often termed guild sovereignty, fails to resolve inherent structural conflicts of interest. Concentrating data ownership, access, and commercialization within any single entity risks inevitably compromising fiduciary duties for profit. Addressing data extractivism requires a distributed fiduciary model that strictly separates control. Through the implementation of legal and technical safeguards such as auditable informational separation of powers, surgical societies can transition into ethical stewards who govern clinical standards while remaining strictly insulated from direct commercial licensing (11).

Finally, this distributed framework relies on legally enforceable reciprocity rooted in the principle of data solidarity (12). The substantial commercial value of AI models emerges from collective clinical aggregation rather than isolated datasets. Therefore, commercial AI developers accessing regional datasets must adhere to rigorous benefit-sharing agreements including mandatory technology transfer, tiered pricing, and the direct reinvestment of algorithm-derived licensing revenues into the public health infrastructures of originating communities. This approach ensures these technological advancements equitably serve rather than exploit the global surgical community.

Acknowledgments

None.

Footnote

Provenance and Peer Review: This article was a standard submission to the journal. The article has undergone external peer review.

Peer Review File: Available at https://hbsn.amegroups.com/article/view/10.21037/hbsn-2026-0206/prf

Funding: This study was supported by the National Natural Science Foundation of China (Nos. 82425049 and 82273074 to T.Y.), the National Science and Technology Major Project of the Ministry of Science and Technology of China (grant Nos. 2024ZD0520500 and 2024ZD0520506 to T.Y.), Shanghai Outstanding Academic Leader Program (No. 23XD1424900 to T.Y.), and Shanghai Health and Hygiene Discipline Leader Project (No. 2022XD001 to T.Y.).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://hbsn.amegroups.com/article/view/10.21037/hbsn-2026-0206/coif). T.Y. serves as an unpaid editorial board member of HepatoBiliary Surgery and Nutrition and reports grants from the National Natural Science Foundation of China (Nos. 82425049 and 82273074), the National Science and Technology Major Project of the Ministry of Science and Technology of China (grant Nos. 2024ZD0520500 and 2024ZD0520506), Shanghai Outstanding Academic Leader Program (No. 23XD1424900), and Shanghai Health and Hygiene Discipline Leader Project (No. 2022XD001). The other authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Varghese C, Harrison EM, O’Grady G, et al. Artificial intelligence in surgery. Nat Med 2024;30:1257-68. [Crossref] [PubMed]
2. Famularo S, Donadon M, Cipriani F, et al. Machine Learning Predictive Model to Guide Treatment Allocation for Recurrent Hepatocellular Carcinoma After Surgery. JAMA Surg 2023;158:192-202. [Crossref] [PubMed]
3. Calderaro J, Seraphin TP, Luedde T, et al. Artificial intelligence for the prevention and clinical management of hepatocellular carcinoma. J Hepatol 2022;76:1348-61. [Crossref] [PubMed]
4. Madani A, Namazi B, Altieri MS, et al. Artificial Intelligence for Intraoperative Guidance: Using Semantic Segmentation to Identify Surgical Anatomy During Laparoscopic Cholecystectomy. Ann Surg 2020;276:363-9.
5. Gorijavolu R, Sakran JV. Who Owns the Operating Room? Surgical Data Governance at a Crossroads. JAMA Surg 2026;161:441-3.
6. Zhang A, Xing L, Zou J, et al. Shifting machine learning for healthcare from development to deployment and from models to data. Nat Biomed Eng 2022;6:1330-45. [Crossref] [PubMed]
7. Chan SL, Sun HC, Xu Y, et al. The Lancet Commission on addressing the global hepatocellular carcinoma burden: comprehensive strategies from prevention to treatment. Lancet 2025;406:731-78. [Crossref] [PubMed]
8. Chagnon CW, Durante F, Gills BK, et al. From extractivism to global extractivism: the evolution of an organizing concept. J Peasant Stud 2022;49:760-92.
9. Rashidian N, Hilal MA. Applications of machine learning in surgery: ethical considerations. Artif Intell Surg 2022;2:18-23.
10. Liu X, Glocker B, McCradden MM, et al. The medical algorithmic audit. Lancet Digit Health 2022;4:e384-97. [Crossref] [PubMed]
11. Erdfelder F, Begerau H, Meyers D, et al. Enhancing Data Protection via Auditable Informational Separation of Powers Between Workflow Engine Based Agents: Conceptualization, Implementation, and First Cross-Institutional Experiences. Stud Health Technol Inform 2023;302:317-21. [Crossref] [PubMed]
12. Prainsack B, Kickbusch I. A new public health approach to data: why we need data solidarity. BMJ 2024;386:q2076. [Crossref] [PubMed]