We are in the midst of an artificial intelligence (AI) revolution. Although AI was first described in the mid-20th century, it was not until recently that its real-world applications started to be realized. Currently, AI powers many of the everyday technologies the world has quickly grown to depend on, including virtual assistants on mobile phones, smart speakers, speech recognition tools for dictation, mobile navigation apps, and chatbots.

Much like its impact on everyday life, AI has the potential to tremendously change health care, including urology. There are numerous opportunities for AI in urology across diverse settings, including clinical care, practice management, education, and research. Clinical care applications are perhaps the most visible use cases for advancing patient care, and applications have been described across the spectrum of clinical settings in urology. The most prominent AI applications include the interpretation of radiographic imaging (eg, prostate MRI, CT scans for nephrolithiasis), evaluation of digital pathology (eg, Gleason grading and diagnosis of prostate cancer), population health applications, and improved prediction models for personalizing clinical care, such as cancer care, based on patient and disease characteristics. In contrast to clinical AI applications, practice management applications do not require clinical validation and may therefore be the earliest applications on the horizon. Such “back-office” applications, including revenue cycle management, automation of prior authorizations, and schedule workflow optimization, are currently under development by several vendors. Importantly, practice management applications may have substantial benefits for physician burnout by reducing administrative workloads. AI applications in other areas, such as surgical education and research, can have important returns as well.

Despite the potential of AI to transform the practice of medicine, there are a number of unique barriers that have slowed its development. Data availability is perhaps the most substantial barrier, as the availability of high-quality, large, curated, and representative health datasets is an essential prerequisite to the development of accurate and generalizable AI models. Clinical applications are also subject to unique challenges, such as the requirement for external validation and regulatory approval. Such requirements will result in longer product development cycles and later deployment than other use cases. Other important barriers include transparency, which is critical for end-user trust and adoption of AI technologies, and practical considerations related to operationalization, implementation, scalability, and cost—all of which have the potential to exacerbate existing health care disparities.

AI in health care also poses unique risks compared with other use cases. Although issues related to copyright violation and data ownership are common to AI in general, the consequences of model error, including bias, can be more serious in clinical applications. In addition, misuse by end-users can have critical ramifications in health care. Both issues raise important medicolegal issues that are unique to AI in health care. Furthermore, given the sensitive nature of protected health information, privacy violation represents a vital concern.

It is evident that policymakers are taking notice of AI in health care. The most widespread legislation has been passed by the European Union. The EU AI Act requires transparency and is protective of fundamental rights, most notably banning facial recognition databases. The EU is likely to pass legislation in 2024 to allow financial compensation to those harmed by AI.¹ The US currently has legislation that coordinates AI use across federal agencies, but as of yet, no comprehensive legislation has been passed. In October 2023, President Biden signed an executive order addressing AI. Within this order, there are disclosure requirements, limitations for development of AI for harmful purposes, and infrastructure assessment. It also permits funding to evaluate AI and notably establishes an AI task force housed within the Department of Health and Human Services with a primary aim of publishing guiding principles for AI in health care.² There is activity in both chambers of Congress. Several initiatives both within committee work and outside committee activities have been developed. These efforts have been bipartisan, with each chamber having dedicated evaluation.³

With this regulatory background in mind, what role can the AUA play in AI health policy and legislation? First and foremost, it is clear that there is a need for physicians from across health care, including urology, to be active in AI health policy discussions. There are several potential strategic initiatives that will help advance health policy efforts, including education on the use and misuse of AI, incorporation of AI technology within each of the AUA committees, and assessing/improving infrastructure for developing and implementing AI technology for urologists. AI will inevitably become an integral part of health care. It is therefore imperative that the AUA and urologists take a proactive role in its evaluation and dissemination by being at the table for the regulatory discussions and assisting the urology workforce in the adoption of AI.