

The House’s One Big Beautiful Bill (HR 1) contains hundreds of pages of changes to tax and health policy. Tucked inside is a small but important provision: $25 million for the Department of Health and Human Services to use artificial intelligence to identify and recover improper Medicare payments.
This is exactly the kind of reform that conservatives should support. It applies new tools to an old problem, avoids creating a new bureaucracy, and signals that Congress is willing to hold government programs accountable. But until Congress reforms how Medicare is structured, no algorithm will be enough to fix the program.
The problem is massive. Medicare pays out more than $800 billion annually. In 2022 alone, nearly $47 billion of that was classified as improper payments. That includes paperwork errors and deliberate scams. Fraud is so common that federal agencies have set up permanent strike forces in cities where criminal billing networks flourish. And those are just the cases we can measure.
Fraud detection today is slow and reactive. Audits take months, and enforcement often comes years after the fact. AI could help. With the right tools, Medicare could scan millions of claims in real time, flag illogical billing patterns, and shorten the lag between fraud and accountability. Private insurers already use these systems, and it’s time the federal government caught up.
AI can reduce the strain on human auditors who are overworked and overwhelmed by the volume of claims. Even when fraud is caught, the process of recovering funds is cumbersome. If AI could front-load fraud detection, it would allow human investigators to focus on the most serious and sophisticated cases, rather than wasting time chasing paperwork errors and clerical mistakes. A hybrid system could pair automated detection with targeted enforcement, improving both speed and precision.
But better detection will not be enough if the system keeps rewarding bad behavior. Medicare’s fee-for-service model pays providers for each test, procedure, and office visit. That model creates a built-in incentive to bill for more. It is also governed by a complex and outdated codebook that few people understand and many can exploit. Honest doctors find it confusing, and dishonest ones find it profitable.
Fixing that problem requires more than software. It requires structural reform.
One path forward is premium support. Instead of paying directly for services, Medicare could offer beneficiaries a fixed contribution to purchase coverage from competing private plans. This would shift the incentive from billing more to delivering value and would also give insurers a reason to invest in fraud prevention and cost control.
Other reforms could help too. Updating the payment structure, reducing the number of billing codes, and aligning payments with outcomes would make fraud harder and oversight easier. In the meantime, Congress could expand the role of recovery audit contractors and encourage state-level innovation. Many states already use AI to track Medicaid fraud. They should be free to build on that progress.
Congress should also reconsider how it defines improper payments. Some are outright fraudulent while others are technical violations of billing policy that may not reflect intentional wrongdoing. AI systems could help differentiate between honest errors and deliberate abuse, but if the rules themselves are vague, enforcement becomes arbitrary. Streamlining Medicare rules could strengthen fraud prevention by making it clearer what is and is not allowed.
The $25 million in the One Big Beautiful Bill is a modest investment with high potential. It suggests that Congress understands the need to modernize oversight and enforce accountability. It also reflects a conservative insight, as government does not always need to grow to become more effective. It sometimes just needs to be smarter with the tools it already has.
Still, conservatives should not mistake innovation for transformation. AI can help Medicare detect waste and abuse, but it cannot eliminate the deeper incentives that encourage it. Technology can reveal the symptoms, but only structural reform can treat the disease.