National Academy of Medicine Launches Historic Initiative to Safeguard Patients in the AI Era

Washington, D.C. — In a defining move for the future of American healthcare, the National Academy of Medicine (NAM) formally launched its "Patient Safety in the Era of AI" initiative on March 4, 2026. This comprehensive national program aims to harness the transformative potential of artificial intelligence to minimize patient harm while establishing rigorous guardrails for the deployment of AI in clinical settings.

The initiative represents a significant evolution in the NAM’s longstanding commitment to healthcare quality, building directly upon the legacy of its landmark 2000 report, To Err Is Human. While that report first elevated patient safety to a national priority, the current landscape—dominated by the rapid integration of generative AI, predictive analytics, and automated decision-making—requires a fundamentally new strategic approach.

On March 3, 2026, a high-level Steering Group convened in Washington, D.C., to inaugurate the effort. The group brought together a diverse coalition of hospital executives, patient advocates, technology leaders, and policy experts. Their mandate is clear but ambitious: to draft a national strategy that leverages AI not just as a tool for efficiency, but as the primary mechanism for closing the "zero harm" gap in U.S. healthcare.

A New Chapter in Clinical Safety

For over two decades, the healthcare industry has struggled to eradicate preventable medical errors. The NAM’s new initiative proceeds from the premise that AI offers the first viable technical solution to overcome the cognitive and systemic limitations that lead to such errors. By analyzing vast datasets to anticipate risks before they materialize, AI systems could theoretically serve as a highly effective safety net for clinicians.

However, the integration of these technologies brings its own set of risks, including algorithmic bias, lack of transparency, and the potential for "hallucinations" in generative models. The initiative seeks to balance these factors, ensuring that the adoption of health AI is both aggressive in its pursuit of safety and conservative in its management of risk.

Gianrico Farrugia, M.D., President and CEO of Mayo Clinic and Co-Chair of the Steering Group, emphasized the gravity of the moment during the launch. "Today marks a pivotal moment in our national journey toward safer care," Farrugia stated. "I am convinced that emerging AI tools may markedly increase patient safety improvements and do so at scale, but realizing that promise requires deliberate, cross-sector collaboration and thoughtful strategy."

Strategic Pillars and Leadership Structure

The initiative is structured to address the multifaceted nature of AI deployment. It will not merely produce a static report but will engage in a dynamic process of drafting strategic domains, soliciting public feedback, and iterating on policy recommendations. The leadership team reflects this cross-sector approach, ensuring that the voices of patients are given equal weight to those of technologists and hospital administrators.

The Steering Group is anchored by three Co-Chairs who represent the critical pillars of the healthcare ecosystem: clinical delivery, patient advocacy, and system administration.

Table 1: Steering Group Leadership and Focus Areas

Wright Lassiter, President and CEO of CommonSpirit Health, framed the adoption of AI not as a choice but as an ethical imperative. "Patient safety is the ideal use case for health AI," Lassiter noted. "Failing to adopt them risks widening existing safety and equity gaps that we already know how to close."

The Roadmap to a National Strategy

The NAM has outlined a multi-phase roadmap for the initiative. Initially, the Steering Group will develop "draft domains" for a patient safety strategy. These domains will likely cover areas such as algorithmic governance, workforce training, data transparency, and liability frameworks.

Following the drafting phase, the initiative will open a public comment period. This transparency is designed to build trust—a commodity that has been somewhat eroded by the opaque nature of "black box" AI algorithms. The final output will be a cohesive national strategic approach that provides health systems with actionable guidelines for procuring, deploying, and monitoring AI tools.

The initiative is supported by a robust coalition of funders, ensuring it has the resources to operate independently of specific corporate interests. Funding partners include the Elevance Health Foundation, the Association for the Advancement of Medical Instrumentation (AAMI), the California Health Care Foundation, and Premier, Inc.

In parallel with the NAM’s strategic work, the Elevance Health Foundation has launched a "Patient Safety Prize." This competition invites innovators to submit transformational AI solutions capable of improving safety outcomes for vulnerable populations, creating an immediate incentive for market-driven solutions to complement the NAM's policy work.

Navigating the Legislative Landscape

The NAM’s initiative arrives at a time of intense legislative activity regarding artificial intelligence. As reported by legal analysts in early 2026, state legislatures across the U.S. have introduced a record number of AI-focused bills. The International Association of Privacy Professionals (IAPP) noted in a March 2026 analysis that state lawmakers are increasingly moving away from broad, omnibus regulations in favor of targeted, sector-specific laws.

Key Trends in State AI Legislation (2026 Session):

• Sector-Specific Focus: A shift toward distinct bills for healthcare, mental health, and employment, rather than one-size-fits-all AI laws.
• Chatbot Transparency: Mandates requiring AI systems to disclose they are non-human, particularly in patient-facing interactions.
• Algorithmic Accountability: Growing demands for "impact assessments" to ensure AI does not discriminate based on race, gender, or socioeconomic status.
• Frontier Model Regulation: States like California and Illinois are introducing strict reporting requirements for the most powerful "frontier" AI models.

This fragmented legislative environment underscores the necessity of the NAM’s national initiative. While states can penalize bad actors or mandate disclosure, they lack the medical expertise to define what constitutes "safe" clinical AI. The NAM aims to fill this void by establishing clinical standards that can inform future legislation and harmonize the disparate state-level rules into a coherent framework for patient safety.

The Human Element in an Automated Future

A central theme of the launch was the preservation of the human element in healthcare. Sue Sheridan, Co-Chair and CEO of Patients for Patient Safety US, highlighted that while AI is often discussed in technical terms, its ultimate value lies in patient empowerment.

"Artificial intelligence is already reshaping care delivery and improving diagnosis, but it is also democratizing information, empowering patients, and leveraging the autonomy of patients," Sheridan explained. Her comments reflect a growing consensus that AI should not replace clinicians but rather augment the patient-provider relationship by removing administrative burdens and reducing the likelihood of human error.

The initiative also plans to address the "trust gap." As AI systems become more autonomous—potentially diagnosing conditions or recommending treatments without immediate human oversight—patients must be assured that these systems are acting in their best interest. The NAM’s strategy will likely include protocols for "human-in-the-loop" verification, ensuring that critical medical decisions always retain a layer of human accountability.

Looking Ahead

As the "Patient Safety in the Era of AI" initiative moves forward, it will face significant challenges. The pace of AI development currently outstrips the speed of scientific consensus. Tools released today may be obsolete in six months, making the creation of enduring safety standards difficult.

However, the consensus among the attendees in Washington was that waiting is not an option. With AI already being deployed in hospital back offices and diagnostic imaging centers, the healthcare system is already in the "Era of AI." The NAM’s initiative is an attempt to ensure that this era is defined not by unforeseen algorithmic harm, but by a historic reduction in medical error.

The Steering Group will continue its work throughout 2026, with public updates and engagement opportunities expected in the coming months. For the tech industry, healthcare providers, and patients alike, the outcome of this initiative will likely define the rules of the road for the next decade of digital health.

Industry Implications

For AI developers and healthcare technology companies, the NAM launch signals a shift toward more rigorous validation requirements. The era of "move fast and break things" is incompatible with clinical medicine. Companies should anticipate that future procurement guidelines from major health systems will align with the standards emerging from this initiative.

Actionable Takeaways for Health AI Developers:

• Prioritize Explainability: Systems that cannot explain their rationale will likely fail to meet upcoming safety standards.
• Equity Audits: Developers must proactively test for bias, as "equity gaps" are a primary focus of the NAM leadership.
• Collaborative Design: Solutions developed in isolation are less likely to succeed than those co-designed with clinicians and patient advocates.

The launch of this initiative confirms that 2026 is the year where health AI transitions from a novelty to a regulated, critical infrastructure. The National Academy of Medicine has planted its flag, signaling that the safety of the patient must remain the north star in a rapidly digitizing world.