America Is Funding AI Chips—But Not the Safety System AI Medicine Needs

May 22, 2026

Washington has already created industrial policy for artificial intelligence. The CHIPS and Science Act directs roughly $280 billion toward semiconductor manufacturing and AI-related research through 2027, including tens of billions in subsidies and tax incentives for chip production and research. Policymakers continue to debate export controls, tariffs, and national security safeguards for advanced AI hardware.

Yet almost none of that effort has gone into the safety infrastructure required for AI-driven medicine — an industry that represents nearly one-fifth of the US economy. If AI systems in medicine underperform or fail publicly, the damage will impact patients, health care providers and investor confidence.

Today, artificial intelligence reads scans, guides surgical robots and flags high-risk patients in hospitals across the country. But the US still lacks the continuous monitoring and evidence systems needed to ensure these technologies remain safe as they evolve.

Aviation confronted a similar challenge decades ago when automation entered the cockpit. The industry responded by building a layered safety architecture around autopilot systems — simulation networks, fleet-wide monitoring and independent accident investigation.

Medicine has built almost none of this.

That gap is not just a health-policy issue. It is an economic infrastructure problem, and one the Department of Commerce is uniquely positioned to address.

Artificial intelligence promises to transform diagnosis, drug discovery and surgical care. But unlike traditional drugs or devices, adaptive medical systems evolve as they learn from new data.

Regulatory models built for static technologies struggle to govern systems that change in real time.

Despite the widespread use of AI diagnostics and robotic surgical platforms in thousands of hospitals—each generating valuable data on performance and outcomes—these institutions largely function as isolated sites rather than as part of a coordinated national learning system. Currently, there is no integrated national infrastructure to collect and unify these signals into a continuous safety system.

Aviation pairs every aircraft’s black box with an investigative system designed to learn from every flight. Medicine is deploying algorithmic black boxes in hospitals without building the institutional system needed to track performance, compare results and intervene when something starts to go wrong.

That absence becomes even more striking when compared with Washington’s broader AI strategy.

Congress and the Commerce Department have rightly focused on semiconductor manufacturing, supply chains and export controls as foundations of American technological leadership. The National Institute of Standards and Technology is expanding work on AI standards and measurement science, including efforts associated with the US AI Safety Institute.

But the scale remains modest and largely general-purpose. Almost none of it is designed around the messy, high-stakes reality of clinical care.

Recent appropriations discussions have steered only tens of millions of dollars toward AI standards and testing, compared with tens of billions directed toward semiconductor competitiveness and AI research more broadly. None of this funding is specifically aimed at building the real-world evidence infrastructure needed to monitor AI-driven medical systems.

AI tools must perform across diverse patient populations, hospital workflows and clinical contexts, where small shifts in data or environment can change how algorithms behave. Without robust monitoring systems, those changes can remain invisible until they produce clinical failures.

The solution is not to slow innovation. It is to build the institutional infrastructure that allows innovation and safety to advance together.

Commerce and the National Institute of Standards and Technology have played this role before. The United States has long relied on measurement science, technical standards and public-private collaboration to make complex technologies reliable at scale — from aviation to telecommunications to semiconductor manufacturing.

NIST’s recent AI Risk Management Framework was designed as a voluntary foundation for organizations across industries. What it does not provide is the sector-specific evidence infrastructure needed for adaptive systems in medicine.

AI-driven medicine requires a similar approach.

Congress should direct and fund Commerce and NIST to build the infrastructure adaptive medical AI requires. Three steps can start the process.

First, Commerce should work with the Department of Health and Human Services to establish a national validation framework for adaptive medical AI. NIST already leads the development of technical standards across multiple industries. Expanding that role to include benchmark testing, performance metrics and audit protocols for AI-driven clinical systems would provide a foundation for safe deployment.

Second, the federal government should invest in a national evidence infrastructure to monitor AI-enabled medical systems across hospitals in real time. The network would collect de-identified data on performance, complications and error rates to create a continuous learning system.

Third, regulators and industry should adopt simulation and stress-testing standards comparable to those used in aviation and financial stress tests. Before widespread deployment, adaptive medical systems should be evaluated against diverse clinical scenarios designed to expose potential failure modes.

Equally important is the need to train professionals capable of operating across the domains this challenge spans. The safe deployment of AI-driven medicine requires expertise in clinical science, engineering, artificial intelligence, regulatory policy and systems analysis. Few academic programs today prepare professionals to work at that intersection — and the US will need many more of them as AI moves deeper into care.

America has faced similar moments before. In the nineteenth century, infectious disease threatened the ships and ports that carried the nation’s commerce. The federal government responded by building quarantine stations and maritime health systems to protect trade.

Public health in the United States began as infrastructure tied to commerce.

AI is already transforming medicine, but Washington has not invested in the safety architecture needed to support it at scale. Just as ports once anchored the nation’s defense against epidemic disease, the country must now build the digital and institutional ports through which the medicine of the future can safely pass.

The choice is straightforward: build that infrastructure deliberately now, or wait for failures to reveal its necessity.

⸻

Author line

Ara S. Khachaturian, PhD, is Founder and Executive Officer of the Brain Watch Coalition. He also serves as Editor-in-Chief of the Journal of Aging Research & Lifestyle, and Chief Scientific Officer of the International Neurodegenerative Disorders Research Center