Pulling off health insurance fraud used to require real skill. Fabricating medical records demanded knowledge of clinical terminology and billing codes. Impersonating a patient or physician over the phone required a human being to make the call. With artificial intelligence, those barriers are largely gone.
A prompt in a large language model can now generate documentation for a procedure that never happened. AI agents can be instructed to call an insurance company thousands of times in a single day without any human involvement. The technology that has streamlined legitimate business operations is being turned against the health insurance industry at scale.
"We believed AI was something that was going to be leveraged against us as an insurance industry for fraud, and now we're starting to see that," said Kurt Spear, vice president of financial investigation and provider review at Highmark.
The warning is not theoretical. In June 2025, the Department of Justice announced its National Health Care Fraud Takedown, the largest in US history, charging 324 defendants including 96 licensed medical professionals across 50 federal districts for schemes involving more than $14.6 billion in intended losses. AI had already made it into the indictments. Two executives of Pakistani marketing companies were charged with using AI to generate fake recordings of Medicare beneficiaries consenting to receive products, which they sold alongside stolen patient data to support approximately $703 million in fraudulent Medicare claims.
Up to $480 billion is lost each year to healthcare fraud, according to the National Health Care Anti-Fraud Association. The Coalition Against Insurance Fraud puts annual US insurance fraud losses across all lines at $308.6 billion, with healthcare fraud estimated at approximately $105 billion annually. Recovery is difficult. Criminal investigations are the primary route to recouping losses, and returns are typically cents on the dollar.
According to the National Health Care Anti-Fraud Association, AI can be used to falsify medical records, create synthetic patient identities, impersonate physicians and scan coverage policies for exploitable gaps. The scale at which the technology operates is what makes it particularly threatening.
"We have customers that have seen 15,000 bot calls in just a couple months," said Jason Barr, vice president of healthcare for Pindrop, an Atlanta-based firm whose voice authentication technology is used by some of the nation's largest health insurers. Pindrop's system runs in the background during calls, analyzing a speaker's voice, cadence and behavior alongside data from carrier signals and devices to determine whether the caller is human.
The quality of synthetic voices has advanced rapidly. About two years ago they were obviously artificial. Today they are far more convincing, and Pindrop's customers have reported AI callers changing accents or mimicking the agent's voice mid-call.
Deepfake medical imaging is an emerging concern. Highmark is deploying a tool capable of detecting anomalies in medical images down to the pixel level, a recognition that the human eye is no longer sufficient in many cases. A study published this year in the journal Radiology found radiologists had only a 75% accuracy rate in distinguishing real from deepfake X-rays. AI also has detectable signatures in written output. Researchers at the University at Buffalo have developed a tool to identify AI-generated radiology reports, finding that large language models tend toward polished, elaborate phrasing while physicians favor concise clinical language.
Federal enforcement is scaling up. CMS and DOJ have outlined plans to replace the traditional "pay and chase" model with a "detect and deploy" strategy, using AI to flag suspicious billing before payments are processed. As part of the 2025 Takedown, the DOJ announced the creation of a Health Care Fraud Data Fusion Center, bringing together the DOJ's Health Care Fraud Unit, the FBI, HHS-OIG and other agencies to deploy cloud computing, AI and advanced analytics against emerging schemes.
False Claims Act recoveries hit a record $6.8 billion in 2025, with healthcare accounting for 84% of total recoveries. The DOJ has also signaled that AI-enabled documentation practices on the provider side, including AI-driven chart reviews and retrospective coding, will face growing scrutiny as potential grounds for False Claims Act liability.
Despite an 87% jump in insurance AI deployments in 2025, fraud losses have not fallen, suggesting that detection investment has not yet kept pace with the sophistication of AI-powered attacks. Carriers that have deployed advanced detection systems report materially better outcomes, but the gap between leaders and laggards is widening.
For carriers that are behind, the practical priorities are becoming clearer. Voice authentication at the call center level is the most immediate gap to close, given the documented scale of bot call activity. Imaging verification tools and AI-generated documentation detection need to be integrated into the claims workflow rather than applied retrospectively. And given the DOJ's stated focus on AI-enabled provider documentation, carriers also need to reassess how they handle AI-assisted clinical records on the intake side — not just as a fraud detection challenge but as a potential liability exposure if inadequate controls are later scrutinized in a False Claims Act investigation.
The member remains a critical line of defense. If a patient receives a statement describing care they never received, that is often the earliest signal that something is wrong. "Some of the best referrals come from members," said Spear.
