Op-ed: Health AI is moving fast. America needs rules that keep up

Healthcare has traditionally been a reactive field, with symptoms appearing, patients seeking care, and diagnosis and treatment following suit. However, advancements in artificial intelligence are changing this paradigm. Wearable devices can detect irregular heartbeats before symptoms manifest, and AI can identify subtle signs of lung cancer on chest X-rays.

Early detection could significantly reduce late-stage cancer diagnoses, making diseases more treatable and survivable. The promise of AI in healthcare is to transition from reacting to disease to predicting and preventing it.

Yet, AI's impact extends beyond the medical field. Forty-one percent of consumers already incorporate AI and social media into their search for healthcare information, with this figure rising to 57% among Gen Z and 53% among millennials. However, adoption does not equate to trust. Among consumers who use open web-based digital assistants for health information, only 33% express very high satisfaction with these tools. Furthermore, 85% of users verify health information found on AI or social media platforms.

To fully realize the transformative potential of AI, it is crucial to ensure that AI systems are safe, trustworthy, and transparent. This requires robust safeguards, smart national oversight, and policies that foster responsible innovation. As AI becomes increasingly integrated into healthcare delivery and consumer actions, the need for appropriate regulations becomes even more urgent.

Currently, policymakers face a choice: they can create a unified, risk-based federal AI framework that safeguards patients while allowing innovators the freedom to develop and deploy life-saving technologies, or they can allow a fragmented patchwork of state laws that could limit access to superior care. This year alone, over 240 health AI bills have been introduced across 43 states.

These bills, if enacted, could force companies to navigate hundreds of different definitions, disclosure rules, and oversight requirements. Without a consistent national framework for AI regulation, this proliferation of bills will hinder the development and deployment of solutions that can expand access to healthcare, facilitate early disease detection, and ultimately save lives.

The consequences of delayed access to AI-enabled healthcare are profound. Over 77 million Americans reside in areas with a shortage of primary care providers, and more than 122 million live in regions lacking sufficient mental health professionals. Digital health tools, remote monitoring, and AI-powered diagnostics are vital in bridging this access gap and enabling early detection. What is needed now is risk-based, smart oversight that encourages innovation rather than stifling it.

Unlike many traditional medical technologies, AI systems can continue to evolve even after deployment. However, our federal regulatory system, designed for static medical interventions such as drugs, vaccines, and pacemakers, is ill-equipped to handle the dynamic nature of AI. Post-market monitoring is essential to track performance declines, unintended consequences, and changes that may affect the tool's efficacy in clinical care.

This scrutiny must be tailored to the application's risk and the vulnerability of the target population, whether pediatric or geriatric. Additionally, performance should be tracked by age and demographic group to identify specific subgroups where generalized metrics may mask failures.

A new standard, developed through a consensus-based process involving over 75 organizations, provides risk-based guidance for organizations monitoring AI performance, detecting model drift, and ensuring greater transparency around performance. While this standard demonstrates the potential for effective oversight, federal policy should build upon these voluntary standards.

Trust in AI is also contingent on the health data that powers these technologies. Modernized, nationally consistent privacy protections are necessary, tailored to the actual risk posed by the data. Currently, at least 20 state laws create a system in which patients' rights vary depending on their location. Washington, Nevada, and Connecticut have passed laws requiring specific opt-in consent before using health data, while most patients assume such consent is covered under HIPAA – a misconception.

Moreover, patients and their doctors need secure access to and sharing of health information across various devices, apps, and healthcare systems. However, technology's potential to improve healthcare is contingent upon patients' ability to access it and the availability of clear reimbursement pathways. Despite digital health solutions' potential to alleviate the burden on an overextended healthcare system, nearly 60% of providers report that these tools are not yet covered by insurance.

This lack of coverage diminishes the impact of digital health solutions. Over the past two years, Congress has observed the growing complexity and fragmentation of state regulations. Every delay in federal action creates additional uncertainty for companies developing AI-enabled health tools and creates barriers for patients and providers who could benefit from these innovations.

As health AI progresses at an unprecedented pace, our policy framework must evolve in tandem. Smart regulation should protect patients while enabling responsible innovation to scale, ultimately paving the way for a healthcare system that predicts and prevents disease rather than merely reacting to it.