Artificial intelligence is already a fact of life for the healthcare sector but for Congress, dealing with it remains a matter for the future.
Congress doesn't appear anywhere close to moving significant legislation. For the most part, lawmakers are stuck in the stage of declaring that they should act, even as businesses embrace the technology and federal agencies issue regulations to address emerging issues and to carry out President Joe Biden's executive order.
Related: CMS outlines limits on Medicare Advantage AI usage
Lawmakers hail the potential of AI while questioning what they should do and how far they should go with legislation.
"Congress has an obligation to encourage the good outcomes from AI and set rules of the road for new innovations to deliver better care for Americans," Senate Finance Committee Chair Ron Wyden (D-Ore.) said at a hearing on AI in healthcare Feb. 8.
"The question is: How do we regulate this?" Sen. Dr. Bill Cassidy (R-La.) said at the same hearing, offering several broad strategies. In September, Cassidy released a white paper detailing some of AI issues and called on healthcare providers to tell him what they need from Congress.
Congress has certainly been looking into AI, not to mention the use of algorithms and machine learning in healthcare settings.
Senate Majority Leader Chuck Schumer (D-N.Y.) held a series of briefings for senators last year on artificial intelligence, including its healthcare applications.
On Feb. 1, Reps. Troy Balderson (R-Ohio) and Robin Kelly (D-Ill.) launched the bipartisan Congressional Digital Health Caucus at a Capitol Hill event focused on AI. Representatives from Amazon, Google, Microsoft, Hippocratic AI and the Consumer Technology Association participated in the briefing.
A handful of significant public hearings have been held: the Finance Committee this month and the House Energy and Commerce Committee in November and December.
Great interest, few bills
Yet health-focused AI legislation remains scarce. A search of bills on Congress' website yields only 14 measures—some of them duplicates—that include the terms "health" and "artificial intelligence." Similarly, a Brendan Center for Justice database includes dozens of AI bills, but only about a fraction include healthcare, mostly in a tangential way.
Wyden sponsored one of those bills, the Algorithmic Accountability Act of 2023, and talked it up at the Senate hearing. The legislation covers healthcare but has a broader reach and instructs the Federal Trade Commission to draft regulations ensuring transparency, privacy, fairness and accountability from companies that use "augmented" or "automated" decision systems. Wyden, Sen. Cory Booker (D-N.J.), and Rep. Yvette Clarke (D-N.Y.) introduced the bill, which so far only has Democratic co-sponsors, in September.
Wyden's bill does get at the sorts of issues lawmakers are asking about in their exploration of AI in healthcare, including several themes that surface repeatedly.
Privacy, along with related security issues, is one of the most common. The Energy and Commerce Committee will start there, Chair Cathy McMorris Rodgers (R-Wash.) said Wednesday.
"We're really focusing on the importance of a data privacy bill, and believing that that's foundational to AI across the board," McMorris Rodgers said.
Another common concern is bias. McMorris Rodgers, for instance, has mentioned the possibility of federal agencies using AI to limit care to older people or people with disabilities.
Others have pointed to studies that found algorithms already in common use disproportionately deny care to Black patients. Dr. Ziad Obermeyer, a University of California, Berkeley School of Public Health professor who found in 2019 that a biased algorithm affected 150 million patients, told the Finance Committee that the tool is still in wide use, despite fixes being available.
One oft-cited goal of any congressional action is ensuring transparency in AI, not just to limit bias but to make sure that it is being used properly, and to guard against abuses such as recent allegations that insurers have been using AI and algorithms to improperly deny care.
Lawmakers are also concerned about how Congress, regulators and the public know that AI programs are in fact working as intended, especially when most people don't understand the technology. Cassidy, for instance, has questioned whether people could verify or validate AI efficacy, whether separate AI systems would have some role and how any of that would get done without compromising companies' proprietary information.
Promoting AI business
One of the reasons for Congress' go-slow approach is a widespread concern about going too far and harming chances for U.S. companies to dominate AI in the future.
"There's such a first-mover advantage," Sen. Mark Warner (D-Va.) said at the Finance hearing, before asking a witness how to get disclosure requirements right.
Exactly what Congress will or can do in an election year when it is already struggling with just passing spending bills and program reauthorizations remains unclear. Schumer tasked committee chairs with crafting AI bills this year, but issues such as national security, election security, deepfakes and keeping ahead of China have gotten more attention than healthcare.
Asked recently if he saw any prospects for legislation emerging from all the talk and hearings this year, Warner said he was most focused on intelligence threats and market manipulation. But even so, he made no promises. "I got some bills," he simply said.
Correction: An earlier version of this story misidentified the state Rep. Robin Kelly (D-Ill.) represents in Congress.
