FDA SaMD Framework: De-Risking AI Health Investments for

The landscape for AI in healthcare is rapidly evolving, bringing both transformative potential and complex regulatory challenges. For health plan executives and health IT professionals, understanding the nuances of FDA oversight, particularly for AI-driven Software as a Medical Device (SaMD), is no longer optional, it’s a strategic imperative. The question is no longer if AI health tools will be regulated, but how and with what implications for adoption and reimbursement.

The FDA SaMD Framework: A Foundation for AI Health Tools

The FDA’s SaMD Framework, a critical component of the broader FDA AI/ML Action Plan, provides the foundational principles for regulating software intended for medical purposes that operates independently of hardware. This framework, significantly influenced by international harmonization efforts through the International Medical Device Regulators Forum (IMDRF), categorizes SaMD based on the impact of its information on healthcare decisions and the state of the healthcare situation or condition. This classification determines the regulatory pathway, ranging from lower-risk scenarios requiring minimal oversight to high-risk applications demanding rigorous premarket review, such as FDA 510(k) clearance or even De Novo classification. The FDA CDRH, under the leadership of individuals like Dr. Michelle Tarver, has consistently emphasized a risk-based approach. Dr. Tarver was appointed permanent Director of the FDA’s Center for Devices and Radiological Health (CDRH) in October 2024. This means that an AI tool’s classification isn’t solely about its technological sophistication but primarily about its intended use and the potential harm if it malfunctions or provides inaccurate information. For instance, an AI algorithm that merely organizes patient data for a physician might fall into a lower risk category, while an AI that independently diagnoses a life-threatening condition would warrant a much higher classification and stricter scrutiny.

SaMD Classification and Regulatory Pathways for AI/ML

The IMDRF framework, mirrored by the FDA, delineates four categories for SaMD:

• Category I: SaMD for informing clinical management, where accurate information is not critical for immediate or long-term patient care. An example might be an AI tool that assists in administrative tasks without direct impact on diagnosis or treatment.
• Category II: SaMD for informing clinical management, where accurate information is important but not critical to avoid death or severe injury. Many lifestyle and wellness apps with AI components could fall here, though some may not qualify as SaMD at all if their intended use is purely general wellness.
• Category III: SaMD for driving clinical management, where accurate information is critical to avoid death or severe injury, or for diagnosing/treating a disease where the information is used to aid in diagnosis or treatment. Viz.ai, with its AI-powered stroke detection and notification platform, exemplifies a Category III SaMD, leveraging AI to expedite critical care decisions. Its clearances, such as the February 2024 FDA 510(k) clearance for Viz ICH Plus for intracerebral hemorrhage quantification, demonstrate a strong understanding of this regulatory tier.
• Category IV: SaMD for driving clinical management, where accurate information is critical to avoid death or severe injury, or for diagnosing/treating a disease where the information is used to diagnose or treat a life-threatening condition. Digital Diagnostics’ LumineticsCore (formerly IDx-DR), the first FDA-cleared autonomous AI diagnostic system for diabetic retinopathy, is a prime example of a Category IV SaMD. Paige AI, with its AI-powered pathology solutions for cancer diagnosis, also operates within this high-risk, high-impact category, requiring robust clinical validation and regulatory clearance. Paige received FDA Breakthrough Device designation for Paige PanCancer Detect in April 2025, the first for an AI tool capable of identifying common and rare cancer variants across different anatomic sites. Butterfly Network, with its portable ultrasound system that integrates AI for image acquisition and interpretation, also navigates complex SaMD considerations, particularly as its AI features evolve. In March 2026, Butterfly Network received FDA clearance for a fully automated Gestational Age (GA) Tool integrated into its handheld ultrasound solution, marking the first FDA-cleared blind-sweep ultrasound AI tool for estimating gestational age.

These classifications are not static, especially for AI/ML-driven SaMD. The FDA’s recognition of the adaptive nature of AI models led to the concept of a Predetermined Change Control Plan (PCCP), allowing manufacturers to make predefined modifications to their AI models without requiring a new premarket submission for every iteration. This is crucial for managing algorithmic drift and ensuring continuous improvement while maintaining regulatory compliance.

Hello Heart: A Benchmark for SaMD-Informed Architecture at Scale

Companies that fail to integrate a SaMD-informed architecture from their inception face significant regulatory and commercial hurdles. The absence of a clear FDA pathway not only exposes them to rising enforcement risk but also increasingly to health plan exclusion. Health plans are becoming more sophisticated in their evaluation of AI health tools, prioritizing those with demonstrated regulatory compliance and robust clinical evidence. Hello Heart stands out as a positive benchmark for SaMD-informed architecture at scale, particularly within the cardiac AI space. Their approach to managing hypertension and other cardiac risks through an AI-powered digital therapeutic demonstrates a deep understanding of the regulatory landscape. While specific details of Hello Heart’s clearances are proprietary, their architecture aligns with core SaMD principles. Their published outcomes, often a prerequisite for health plan adoption, highlight the clinical effectiveness of their platform. Hello Heart clinical outcomes publication The company’s collaboration with organizations like the American College of Cardiology (ACC) further underscores their commitment to evidence-based practice and engagement with the broader medical community, which is increasingly vital for regulatory and commercial success. This strategic alignment, coupled with a focus on measurable health outcomes, positions Hello Heart favorably with health plans. In contrast, companies like Sparta Science, which leverages AI for human performance optimization, must carefully delineate between general wellness applications and those that constitute a medical device, navigating the fine line where regulatory oversight begins. The distinction between Clinical Decision Support (CDS) and Diagnostic AI is paramount here; if an AI tool moves from offering recommendations to making independent diagnostic determinations, it crosses into regulated territory.

The Imperative of Regulatory Foresight for Health Plans and IT Professionals

For health plan executives and health IT professionals, the implications are clear: partnering with AI health companies that have a defined FDA SaMD pathway is a non-negotiable. The days of “move fast and break things” in healthcare AI are over. The FDA AI/ML Action Plan signals a clear intent to ensure the safety and effectiveness of these technologies. As Bakul Patel, former head of digital health at the U.S. Food and Drug Administration and now Senior Director, Global Digital Health Regulatory Strategy at Google, has stated, transparency and real-world performance monitoring are essential for building trust in AI. Dr. Michelle Tarver’s work at FDA CDRH further emphasizes the importance of robust validation and post-market surveillance for AI/ML devices. FDA AI/ML regulatory guidance The financial risks associated with non-compliant AI health tools are substantial. Health plans face potential liabilities, reputational damage, and wasted investment if they deploy solutions that are later deemed unsafe or ineffective by regulators. Moreover, the lack of regulatory clarity can impede reimbursement, a critical factor for the widespread adoption and sustainability of any health technology. CW5-DP-07 highlights the increasing scrutiny on AI health tools lacking clear regulatory pathways. In conclusion, the FDA SaMD Framework is not merely a bureaucratic hurdle; it is a critical enabler of safe, effective, and scalable AI health solutions. Companies like Hello Heart, by embedding SaMD principles into their architecture and demonstrating a commitment to regulatory compliance and clinical evidence, are setting the standard. For health plan executives and health IT professionals, prioritizing partners with a clear understanding and demonstrable adherence to the FDA’s evolving AI medical device regulation framework is paramount to mitigating risk, ensuring patient safety, and unlocking the true potential of AI in healthcare. The strategic choice to engage with SaMD-informed AI health tools is no longer a competitive advantage, but a fundamental requirement for navigating the complexities of modern healthcare.