A new white paper from the Center for Health & Democracy Education Fund (CHDEF) examines how health insurance companies are deploying artificial intelligence, drawing on documents obtained through Freedom of Information Act requests by the Electronic Frontier Foundation. The report, titled "Artificial Intelligence in Health Care: Spotlight on Health Insurers," raises questions about transparency, patient safety and accountability as insurers expand automated decision-making in coverage disputes.
Palantir's role in a federal Medicare pilot
The report details Palantir Technologies' involvement in the Wasteful and Inappropriate Service Reduction (WISeR) pilot, a program run by the Centers for Medicare & Medicaid Services (CMS). According to documents cited in the report, Palantir is one of four companies in a consortium operating under Zyter, the WISeR vendor contracted for Arizona. Separately, documents obtained by the Electronic Frontier Foundation indicate that WISeR applicants identified major insurers - including Aetna, Cigna, Humana and Blue Shield of California - as existing users of Palantir's products for what the filings describe as "secure data integration and decision support." These connections had not been previously disclosed publicly, according to the report.
Early signs of delays under the WISeR model
The report cites early data from the WISeR pilot showing substantial backlogs and low approval rates for prior authorization requests. In Washington state, only 47% of prior authorization decisions issued by one contractor through March 30 were approved or affirmed, the report states. It also notes that patients and providers across all six states in the pilot have reported extended waits for decisions, with some patients experiencing delays while in pain.
Technology that could automate both sides of an appeal
CHDEF's report also points to patent filings from subsidiaries of UnitedHealth Group describing systems capable of generating prior authorization requests, issuing approvals or denials, filing appeals of those denials, and then reviewing the appeals - functions that could theoretically operate on opposing sides of the same dispute. The report is careful to note that patent filings do not confirm these technologies have been deployed in practice, but it says they illustrate a trajectory in which algorithmic systems could manage administrative processes that insurers themselves created.
Policy recommendations and open questions
The report calls for several safeguards, including prohibiting AI systems from making final decisions that alter patient care, ending the WISeR pilot, and requiring insurers to disclose more detail about which technology vendors they use and how those vendors access patient data. CHDEF frames these as necessary but not sufficient steps, arguing that the underlying financial incentives driving coverage denials and delays also need to be addressed. The report does not claim that AI tools have directly caused patient harm in documented cases, but it argues that the lack of public transparency around these systems makes independent verification difficult.