President Donald Trump wants to speed adoption of AI by enabling regulatory sandboxes where researchers, startups and enterprises can deploy and test AI tools.
The idea was raised in “America’s AI Action Plan” that the administration released in July. The action plan promotes American dominance in the creation and adoption of AI across sectors.
The healthcare sector was noted to be a slow adopter of technology because of “distrust or lack of understanding of the technology, a complex regulatory landscape, and a lack of clear governance and risk mitigation standards,” the action plan says.
The Trump administration proposed a series of AI Centers of Excellence where model developers could try out the technology without fear of backlash from federal agencies. For healthcare, the Food and Drug Administration (FDA) would enable AI Centers of Excellence, supported by the AI evaluation initiatives at the National Institute of Standards and Technology.
James McHugh, managing director at Berkeley Research Group, said that healthcare is not yet at a point where health systems can clearly invest in one technology alone. Because AI is still use case-driven, hospitals still need to do a trial run for new models. “The capabilities are coming so quick, and we have to figure out a way to react to it appropriately and not overinvest,” he said.
Regulatory sandboxes would be useful to test drive new technologies.
“The reason the sandbox is really important, it allows you to experiment with a number of different platforms in that way, instead of over-committing to it, so it's perfectly aligned with where I think the industry is right now,” McHugh said.
The implementation of the AI Centers of Excellence could be difficult, McHugh said. The AI Action Plan doesn’t propose many details for the centers, but McHugh suggested the centers be managed at a national level, rather than regionally. Because local providers are competitive with one another, they may not want to engage in new product development if the COEs were regional.
It may also be difficult to encourage collaboration between organizations because each system has its own individual needs and priorities.
McHugh also fears that smaller organizations could be left behind in the COE model because it is opt-in. Large organizations that are already engaged in AI would be more likely to opt in than small providers with limited to no knowledge of the technology. However, he hopes that the learnings from the COEs could be distributed to less-resourced organizations.
“When some of the regulations were being tossed around, it was going to be too strict, and it might be a little too limiting,” McHugh said. “So the idea is that it's opening up the space to be a little bit more collaborative, and then it's actually forcing accountability back on the providers to come up with their own structure too.”
BRG had already been encouraging its clients to stand up internal centers of excellence for AI rooted in the IT department, so that its governance could be centralized. What the consulting group found is that IT departments can be reluctant to take on the role because of its increasing budget size.
Healthcare is in a cost-cutting phase because of Trump’s One Big Beautiful Bill Act, which cuts Medicaid eligibility, and research grant cuts that affect academic medical centers. McHugh is urging clients to see that the only way to find more efficiency in the budget is to use technology.
“My last pitch to the providers is, you have to demonstrate that you're ready to [participate], that you have a willingness, a readiness, and you have an oversight structure in place,” McHugh said. “So, in order to participate in the sandbox or Center of Excellence, you really kind of need your own.”