policy
Debate on state AI laws heats up
I spent yesterday a stone's throw from the White House at a health care AI summit headlined by deputy health secretary Jim O'Neill. I wrapped up a number of his comments in a story, but most provocative was what he didn't say: When event host Joe Grogan asked about a controversial proposal to prevent states from enforcing their AI laws, O'Neill demurred, saying only that in general he supported fighting state regulation on AI.
For the last few weeks a draft executive order that would take various actions to undermine state laws AI has been circulating. And, in a Truth Social post this week, President Trump indicated some order to this effect was imminent. Maybe O'Neill just didn't want to steal the President's thunder.
But the idea that states shouldn't make their own rules in the absence of federal leadership on regulation is controversial. Multiple congressional efforts to pass legislation that would block states' ability to enforce local laws failed this year owing to stiff opposition on both sides of the aisle — even some very Republican states have passed their own AI laws.
During the event I chatted with Zach Boyd, who directs Utah's Office of Artificial Intelligence Policy where he led a successful effort to pass a law concerning mental health chatbots. Obviously, he does not want to see his work undone. During comments on a panel he allowed both that "there are some things at the state level that we really should be doing," like handling issues of consumer protection, and that "there are some issues where I would really welcome federal preemption," for example on issues of national security. Among other things, Boyd is skeptical of claims that over 1,000 state laws regulating AI have been proposed or passed.
"This 1,000 bill number has been much touted in various circles, and it is true, there are 1,000," he said. "Whether they're mostly regulatory, I actually think that factually, that's not true."
Part of it comes down to your definition of regulation, said Boyd. "I think most of them are doing things like establishing a state AI task force, or setting rules for government use of AI, or clarifying criminal codes around deep fake AI."
overheard
Will interoperability ever happen?
In the same health AI summit, heavy hitters weighed in on the prospects for interoperability, or the effort to make health data currently scattered about the health care system more readily accessible to patients, their providers, and maybe even researchers. A few comments of note:
- This was news to me: Thomas Keane who is the top health IT regulator at HHS said that before Christmas his office hopes to introduce a rule that would eliminate 34 certification criteria for electronic medical records and modify seven. These certification criteria have been much maligned for various reasons including that they create a huge burden on small companies that might want to compete with dominant vendors like Epic Systems.
- "We would like the EHR companies to have to do less conformance testing and concern themselves less with this aspect of our regulatory regime," said Keane. "What we will pivot towards in the spring is regulating the application programming interfaces by which the EHRs talk to each other. And we can say that you have to adopt these features, you have to support this functionality, and if that functionality exists, then these EHRs and other technological applications in the ecosystem should all be able to talk to each other."
- Investor Vijay Pande has resurfaced following his exit from Andreessen Horowitz's Bio & Health team earlier this year. He now has a new healthcare AI fund called VZ.VC that he's not really ready to talk about yet, even though it's on his LinkedIn profile.
- Pande offered an intentionally provocative take on the prospects for interoperability: "I think it's never going to happen. I think there's not enough market forces to drive that... I think there's reasons why we haven't seen it — It's not the technology."
medical devices
Breakthrough devices appear to overcharge CMS
So what happens when you enable special payments for devices deemed breakthroughs by the Food and Drug Administration? An analysis of hospital pricing in JAMA Health Forum suggests many devices taking advantage of special CMS' special payments quote higher prices to the government insurer than they ordinarily charge hospitals. "These findings could indicate that some manufacturers overstate device costs to trigger or augment supplemental Medicare reimbursement for hospitals," the authors write.
Importantly, breakthrough devices are able to receive marketing authorization from FDA when the regulator has greater uncertainty about their risks and benefits. You can read all about that in the investigation into the program that Katie Palmer and I did several years ago.
No comments