The Blue Cross Blue Shield Association says hospitals' use of AI tools to prepare insurance claims added about $942 million in healthcare spending over two years, TechCrunch reported Saturday, citing an analysis first described by The New York Times. The association argues the tools pushed up the complexity of recorded diagnoses without any matching change in the care patients received.

What Happened

BCBSA, the national federation of Blue Cross and Blue Shield insurers, examined claims that hospitals submitted with the help of AI tools over a two-year period. It found a sharp increase in the number of patients documented as having complex conditions, which typically command higher reimbursement. According to the association, there was a clear disconnect between how patients were coded and how they were treated: the documentation became more severe, while the care delivered did not.

The Times framed the analysis as further evidence that AI is already raising healthcare costs, and noted that both hospitals and insurers are now deploying AI in their long-running disputes over payment.

RelatedNew York City Comptroller Urges Anthropic to Delay IPO Over Safety and Governance Concerns

Dr. Shiv Rao, founder of the AI clinical documentation startup Abridge, acknowledged the risk of a future in which competing systems simply battle each other, which he described as bots fighting bots and agents fighting agents. He argued that AI could instead reduce tension between the two sides and bring costs down. Luke Chalker, a senior vice president at BCBSA, rejected the idea of an even contest, saying insurers are on the losing end of what he called a completely "one-sided blood bath."

Why It Matters

AI scribes and coding assistants have spread quickly through U.S. hospitals because they reduce paperwork for clinicians and capture details from patient encounters that might otherwise go unrecorded. Under fee-for-service and risk-adjusted payment models, more detailed diagnoses can translate directly into higher payments. Hospitals and vendors say that reflects care that was already being provided but poorly documented; insurers call the pattern upcoding.

The BCBSA figure gives insurers a concrete number to put in front of employers, regulators and lawmakers. It also puts documentation vendors in an awkward position, since their pitch to hospitals often rests on improved capture of billable work, while their pitch to the wider system rests on lower administrative costs.

The analysis has limits worth noting. It comes from a party to the dispute, and TechCrunch's report did not include responses from hospitals or AI vendors to the specific findings. A two-year rise in coded complexity can have several causes, and the association's argument depends on its claim that treatment patterns stayed the same.

Context

Friction between providers and payers over billing codes predates AI by decades, with each side accusing the other of gaming the rules. What has changed is automation. Both hospitals and insurers can now process claims and denials at a volume and speed that human coders and reviewers could not match, which raises the stakes of every rule change and audit.

What to Watch

Insurers may lean further on their own AI to review and contest claims, and hospitals may respond with tools built to rebut denials. Watch for whether federal or state regulators, or large employers that fund health plans, ask for audits of AI-assisted coding, and whether hospitals or vendors publish data challenging the association's conclusion that care did not change.