AI is expected to reduce healthcare costs and improve efficiency, but reality is moving in the opposite direction. According to a newly released analysis by the Blue Cross Blue Shield Association (BCBSA), hospitals are increasingly using AI tools when submitting insurance claims,driving up healthcare spending by an additional $942 million over two years—while patients receive no additional treatment.
Core data: Where did the $942 million come from?
BCBSA’s analysis notes that hospitals are increasingly relying on AI tools to “optimize” diagnostic coding when submitting claims to insurers. The result is:A sharp rise in the number of patients recorded as having “complex conditions”Complex conditions trigger higher reimbursement rates, so claim amounts rise accordingly.
But the problem lies in a sharply worded term the report uses to describe this misalignment—“a clear disconnect between coding and treatment”In other words, “complex conditions” on paper are proliferating, yet there is no evidence that patients’ actual delivered care has correspondingly changed. Put simply, AI helps hospitals write thicker bills—but not more substantial treatment.
The “AI arms race”: Both providers and payers ramp up AI use
The New York Times reports that this incident is merely the latest signal of AI inflating healthcare costs—and the real trouble is:Hospitals and insurers alike are deploying AI, fueling a mutually escalating “arms race”.
- On the hospital sideAI is used to optimize coding and maximize claim amounts—upcoding cases to higher-reimbursement categories wherever possible.
- On the insurer sideAI is used to audit claims, detect “upcoding,” suppress reimbursements, and even automatically deny claims.
Both sides deploy AI; both sides “optimize” their own interests—resulting in higher system friction costs: more audits, more disputes, more administrative overhead. These costs ultimately get passed on through premiums, landing on every policyholder.
It’s not AI’s fault—it’s a problem with the “objective function”
To be clear: The issue lies not with “AI technology” itself, but withthe objective being optimizedWhen AI is trained as a tool to “maximize payout amounts” or “minimize payouts,” it will inevitably exploit loopholes in the rules.
This is actually a common challenge across all AI applications:AI excels at optimizing the objectives you set—but if those objectives themselves are flawed, AI will amplify that flaw to the extremeHealthcare is merely one of the earliest domains to expose this issue—similar dynamics are already playing out in finance, advertising, insurance claims processing, and even content recommendation. This “objective misalignment” is also one of the root causes of AI safety controversies,OpenAI’s dissolution of its Safety Preparedness teamreflects the same tension between commercial goals and safety responsibilities.
What this means for ordinary people
For patients and policyholders, this situation carries several practical implications:
- Premiums may continue risingRising claims friction and administrative costs will ultimately be passed on to everyone via premiums.
- “AI-denied claims” riskAs insurers deploy AI for automated claims review, the likelihood of erroneous denials increases—and patients may need to proactively appeal.
- Beware of “overcomplication”If your bill includes an unfamiliar “complex diagnosis,” it’s worth asking for clarification.
This “AI vs. AI” arms race is unlikely to end anytime soon. Unless the industry reaches consensus on coding standards and claims review rules, AI adoption in healthcare will only increase surface-level complexity—not necessarily real medical value.
How regulators and the industry will respond
This contest has already drawn regulatory attention. BCBSA’s report itself signals the insurance industry’s intent—to bring the issue of “hospitals overcoding with AI” into public and regulatory view—and to push for stricter coding audit standards.
Meanwhile, hospitals argue that AI merely helps them record diagnoses more accurately and avoid underreporting—not intentionally inflate claims. At its core, this dispute is the inevitable result of AI amplifying a long-standing “gray area” in medical coding: AI did not create the problem; it only made existing issues more visible and faster-acting.
It is foreseeable that as AI spreads across both ends of the healthcare claims process, regulators will eventually intervene—either by standardizing coding review criteria or by imposing transparency and auditability requirements on AI tools. Until rules become clear, ordinary patients can only stay more vigilant about their medical bills and ask one more question: “Why?”
Frequently Asked Questions (FAQ)
Is AI really making healthcare more expensive?
According to BCBSA’s analysis, hospitals’ use of AI to optimize claims coding drove approximately $942 million in additional spending over two years. Note, however, that this reflects growth in “claims amounts,” not growth in “healthcare services”—the report identifies a clear disconnect between the two.
Why are both hospitals and insurers adopting AI?
Hospitals use AI to “upcode” cases into higher-reimbursement categories and boost revenue; insurers use AI to detect overcoding and suppress payouts. Though their goals oppose each other, the outcome is rising administrative friction costs across the entire system.
Will this affect China’s medical insurance system?
This case occurred within the U.S. commercial insurance system, but the underlying logic—AI optimizing coding and AI auditing claims—could unfold across global healthcare systems. China’s medical insurance system is also advancing AI-based auditing, and patients must similarly pay attention to issues like “AI-driven claim denials” and “overcoding.”
What other applications does AI have in healthcare?
Beyond claims coding, AI is applied in medical imaging diagnostics, drug discovery, and video-based consultations. For example, Google’s AMIE medical AI has already achieved video consultation capability; details are available in our related coverage.
Want to understand AI’s real-world impact across industries? Explore our AI Model Library and Tool Comparison Engineor continue reading:All review articles · AI Enters the Military Command Chain · Why AI Giants Are Collectively Slowing Down · Three Major AI Giants Jointly Establish SAFA.