Healthy Technology Act

Overview

The Healthy Technology Act of 2025, introduced as H.R. 238 in the 119th Congress, represents a significant legislative proposal aimed at modernizing the United States healthcare regulatory framework to accommodate autonomous artificial intelligence (AI) and machine learning (ML) technologies. Introduced by Representative David Schweikert on January 7, 2025, the bill seeks to amend the Federal Food, Drug, and Cosmetic Act (FDCA) to formally recognize AI and ML systems as 'practitioners' capable of prescribing medications. This move is designed to address systemic inefficiencies in the healthcare system, such as physician burnout and limited access to care in rural or underserved areas, by allowing validated technological solutions to perform clinical functions traditionally reserved for human professionals. The legislation is a reintroduction of concepts previously explored in earlier congressional sessions, reflecting a growing legislative interest in the integration of high-level AI into direct clinical practice. By targeting Section 503(b) of the FDCA, the bill creates a legal bridge between federal medical device regulation and state-level medical practice laws. It does not grant blanket authority to all AI systems; rather, it establishes a strict dual-authorization pathway that requires both federal safety clearance and state-level statutory recognition. As the healthcare industry increasingly shifts toward digital health and data-driven diagnostics, H.R. 238 stands as a foundational attempt to define the legal status of AI as an independent actor in the medical prescription process. The bill's introduction comes at a time when the FDA is already grappling with the rapid advancement of Software as a Medical Device (SaMD), and H.R. 238 provides the necessary legislative clarity to move beyond decision support into autonomous action.

Definitions and Legal Scope

The primary legal contribution of H.R. 238 is the expansion of the term 'practitioner licensed by law to administer such drug' under Section 503(b) of the Federal Food, Drug, and Cosmetic Act (21 U.S.C. 353(b)). Historically, this definition has been interpreted to include only human medical professionals, such as doctors, dentists, and authorized nurses, who are licensed by their respective states. The bill explicitly amends this subsection to include 'artificial intelligence and machine learning technology,' thereby creating a new category of non-human practitioners within the federal legal code. This definition is critical because it allows for the legal recognition of prescriptions generated by algorithms, provided they meet the subsequent safety and authorization criteria. Furthermore, the bill defines the technical scope of eligible AI/ML technologies by referencing specific sections of the FDCA related to medical device oversight. To qualify as a practitioner, the technology must be approved, cleared, or authorized under sections 510(k) (premarket notification), 513 (device classification), 515 (premarket approval), or 564 (emergency use authorization). By tying the definition of an AI practitioner to these established FDA pathways, the bill ensures that only technologies that have undergone rigorous clinical evaluation and safety testing can be granted prescriptive authority. This technical definition prevents the use of unregulated or 'black-box' algorithms in patient care, ensuring that AI practitioners are held to the same high standards as physical medical devices. The inclusion of Section 564 is particularly noteworthy, as it allows for the use of AI practitioners during public health emergencies, potentially providing a scalable solution for medication distribution during pandemics or other crises where human medical resources are severely constrained.

Governance and Institutional Framework

The governance framework established by the Healthy Technology Act of 2025 is a bifurcated model that balances federal oversight with state sovereignty. At the federal level, the Food and Drug Administration (FDA) serves as the primary technical regulator. The FDA is responsible for evaluating the safety, efficacy, and performance of the AI/ML algorithms through its existing medical device review processes. This includes assessing the technology's ability to accurately diagnose conditions and determine appropriate pharmacological interventions. The FDA's role is focused on the 'product' aspect of the AI, ensuring that the software is robust, secure, and free from significant bias that could lead to medical errors. The agency must ensure that the AI's decision-making process is transparent enough for regulatory audit and that the data used to train the model is representative of the diverse patient populations it will serve. At the state level, the bill respects the traditional role of state governments in regulating the 'practice of medicine.' For an AI system to legally prescribe drugs, it must be authorized pursuant to a statute of the state involved. This means that federal FDA clearance is a necessary but not sufficient condition for deployment. State legislatures must pass specific laws that recognize AI as an authorized practitioner within their jurisdiction. This institutional framework ensures that states can implement their own oversight mechanisms, such as requiring AI systems to be registered with state medical boards or limiting the types of drugs (e.g., non-controlled substances) that an AI is permitted to prescribe. This dual-layered approach prevents federal overreach while providing a clear pathway for technological adoption.

Key Focus Areas and Clinical Applications

The central focus of H.R. 238 is the legal enablement of autonomous prescriptive authority for AI and machine learning technologies. By focusing on Section 503(b) of the FDCA, the bill specifically addresses 'prescription-only' drugs—those that require professional medical supervision due to their toxicity, potential for harmful effects, or the complexity of their use. The bill aims to ensure that AI can provide the necessary level of supervision by processing patient data, medical histories, and real-time physiological inputs to make safe prescribing decisions. This focus is particularly relevant for chronic disease management, where AI can provide continuous adjustments to dosages that would be impractical for human physicians to manage on a daily basis. For example, an AI system could manage insulin dosages for diabetic patients or adjust hypertension medications based on wearable sensor data. Another key focus area is the integration of AI into the existing pharmaceutical and healthcare IT infrastructure. The bill seeks to ensure that prescriptions generated by authorized AI systems are treated with the same legal validity as those issued by human doctors. This requires alignment with electronic prescribing (e-prescribing) standards, pharmacy management systems, and Electronic Health Records (EHRs). By focusing on the operational legality of the AI-generated prescription, the bill aims to create a seamless workflow where a patient can receive a diagnosis from an AI system and have the prescription transmitted directly to a pharmacy without the need for a human intermediary to 'co-sign' the order, thereby maximizing the efficiency gains of the technology and reducing the administrative burden on the healthcare system.

Implementation Framework and FDA Pathways

The implementation of the Healthy Technology Act of 2025 follows a two-stage process involving both federal and state actions. For technology developers, the implementation begins with the FDA's premarket review process. Developers must submit comprehensive data demonstrating that their AI/ML models can safely perform the functions of a prescribing practitioner. This involves clinical trials or real-world evidence showing that the AI's prescriptive accuracy is at least equivalent to that of a human clinician. The implementation framework also anticipates the use of the FDA's 'Total Product Lifecycle' approach, where the technology is monitored for performance drift and updated through established change control protocols to ensure ongoing safety. The FDA may require specific 'Predetermined Change Control Plans' (PCCPs) that outline how the AI will learn and adapt after it is deployed in the field. The second stage of implementation occurs at the state level, where developers must navigate the legislative and regulatory requirements of individual jurisdictions. States may choose to implement the Act through various models, such as pilot programs, specialized licensing for AI developers, or the creation of new oversight committees within state health departments. The implementation framework also necessitates updates to healthcare provider workflows, as hospitals and clinics integrate AI practitioners into their care teams. This includes establishing protocols for human-AI collaboration, where human physicians may still provide high-level oversight while the AI handles routine or data-intensive prescribing tasks. The bill does not mandate the use of AI but provides the legal infrastructure for those healthcare systems that choose to adopt it.

Monitoring, Evaluation, and Post-Market Surveillance

Monitoring and evaluation under H.R. 238 are primarily conducted through the FDA's post-market surveillance and adverse event reporting systems. Because AI and machine learning systems are dynamic and can evolve as they process more data, continuous monitoring is essential to detect any degradation in prescriptive accuracy or the emergence of unforeseen risks. The bill relies on the FDA's existing authority to require manufacturers to report 'malfunctions' or 'adverse events' associated with their devices. In the context of an AI practitioner, this would include any instances where the AI prescribed an inappropriate medication or dosage that resulted in patient harm, triggering a regulatory review of the system's authorization. Evaluation also extends to the broader impact of AI practitioners on the healthcare ecosystem. Policymakers and researchers will likely evaluate the technology's effect on patient outcomes, healthcare costs, and provider workloads. This includes assessing whether AI prescribing reduces the time-to-treatment for patients in underserved areas and whether it effectively mitigates the administrative burden on human doctors. The evaluation process is intended to be iterative, with the data collected from early deployments informing future regulatory refinements and the development of best practices for AI-driven clinical care. This ensures that the recognition of AI as a practitioner remains grounded in empirical evidence of its safety and utility. Furthermore, the FDA may utilize its Sentinel System to proactively monitor the performance of AI practitioners across large-scale electronic health data, identifying trends that may not be apparent through individual adverse event reports.

Penalties, Liability, and Tort Implications

The Healthy Technology Act of 2025 operates within the established enforcement framework of the Federal Food, Drug, and Cosmetic Act. AI developers who deploy prescriptive technologies without the required FDA clearance and state authorization face significant penalties, including civil fines, injunctions, and the potential for criminal prosecution for the distribution of 'misbranded' or 'adulterated' medical products. Furthermore, if an AI system is found to be operating outside the scope of its authorization—such as prescribing drugs for which it was not cleared—the FDA has the authority to issue recalls or revoke the system's practitioner status, effectively shutting down its clinical operations. Liability remains one of the most complex legal issues associated with the bill. By recognizing AI as a 'practitioner,' the legislation raises questions about medical malpractice and product liability. If an AI system makes a prescriptive error, the legal system must determine whether the developer, the healthcare facility, or a supervising physician is responsible for the resulting harm. While H.R. 238 does not explicitly define these liability shifts, it sets the stage for state courts and legislatures to develop new tort frameworks for AI-driven medicine. This may include the evolution of the 'learned intermediary' doctrine, which traditionally shields manufacturers from liability if they provide adequate warnings to a prescribing physician. If the AI is the prescriber, the manufacturer may face direct liability for the algorithm's decisions. Appeals processes for developers whose systems are denied authorization follow the standard administrative procedures of the FDA and state medical boards, providing a pathway for legal challenge and review of regulatory decisions.

Relationship to Other Federal Instruments

H.R. 238 is designed as a targeted amendment to the Federal Food, Drug, and Cosmetic Act (FDCA), specifically 21 U.S.C. 353(b). It does not replace the existing regulatory structure but rather integrates AI into the established legal definition of a practitioner. This relationship ensures that the new AI practitioners are subject to the same general requirements as human practitioners regarding the labeling and distribution of drugs. However, the bill also interacts with the 21st Century Cures Act, which established the initial framework for regulating software as a medical device (SaMD). H.R. 238 represents the next logical step in this regulatory evolution, moving from software that supports clinical decisions to software that autonomously executes them. The bill also has important implications for the Controlled Substances Act (CSA), administered by the Drug Enforcement Administration (DEA). While H.R. 238 focuses on the FDCA, the recognition of AI as a practitioner may eventually require the DEA to update its regulations to determine if AI systems can hold DEA registrations for prescribing controlled substances, which are subject to much stricter oversight than non-controlled medications. Additionally, the Act must be reconciled with the Health Insurance Portability and Accountability Act (HIPAA), as AI practitioners will necessarily process vast amounts of Protected Health Information (PHI). The relationship between these various federal instruments creates a comprehensive regulatory environment that governs the technical, clinical, and privacy-related aspects of AI in healthcare, ensuring that the introduction of autonomous technology does not compromise patient privacy or public safety.

International Alignment and Global Standards

The Healthy Technology Act of 2025 positions the United States as a leader in the regulation of autonomous medical AI, but it also reflects broader international trends toward the oversight of high-risk AI systems. For example, the European Union's AI Act classifies AI systems used in medical devices as 'high-risk,' requiring them to meet strict standards for transparency, data quality, and human oversight. While the EU framework focuses heavily on conformity assessments and risk management, H.R. 238 goes a step further by granting AI a specific legal status (practitioner) that is unique to the U.S. healthcare context. This U.S. approach emphasizes the role of sub-national (state) authorization, which contrasts with the more centralized regulatory approach of the EU. Despite these differences, the bill aligns with international technical standards developed by organizations like the International Medical Device Regulators Forum (IMDRF). By utilizing the FDA's 510(k) and PMA pathways, H.R. 238 ensures that AI practitioners are evaluated using criteria that are recognized globally. This alignment facilitates the potential for future mutual recognition agreements, where an AI system authorized in the U.S. could be more easily evaluated for use in other jurisdictions that follow similar SaMD guidelines. As AI-driven healthcare becomes a global phenomenon, the Healthy Technology Act provides a model for how nations can integrate autonomous technologies into their domestic legal systems while maintaining high standards for patient safety. The bill's focus on clinical performance and state-level licensing provides a pragmatic framework that other federalized nations may look to emulate as they grapple with the legal status of autonomous medical software.

Implementation Timeline

Compliance Checklist

Sources and References