The Medicare AI Deciding Care Denials Operates in Deliberate Darkness
The EFF's FOIA lawsuit against CMS forces transparency on WISeR, an AI care-denial program affecting millions of seniors with no public documentation.
An Algorithm Makes the First Cut on Senior Care — and Nobody Outside CMS Knows How
WISeR did not begin as a policy debate. It began as a January 2026 deployment, launched into a federal healthcare infrastructure serving tens of millions of Medicare beneficiaries, with no accompanying documentation about how the AI's decisions are reached, what data it was trained on, or how often it gets the determination wrong. The FOIA lawsuit demanding program records that the EFF filed against CMS is, at its core, a demand that the agency produce the paper trail that should have accompanied any program affecting this many people at this level of consequence.
Prior authorization is already the most contested chokepoint in American healthcare — the place where administrative processes most visibly intersect with clinical judgment. Inserting an AI at that chokepoint without public disclosure of its operating principles is not a technical choice; it is a governance choice. CMS made that choice in January and has not reversed it.
Opacity as Policy: Why CMS's Silence Is the Decision, Not the Delay
Federal agencies deploy new programs without full public disclosure routinely — the question is whether the scope and consequence of WISeR places it in a category that requires more. The EFF's argument, embedded in its FOIA filing , is that it does: a program that determines whether seniors receive care is not an internal administrative process exempt from scrutiny, it is a public-facing adjudication system whose criteria belong in the public record.
The absence of those criteria is not a gap to be filled later. For every Medicare beneficiary who received a denial through WISeR between January and March 2026, the AI's reasoning is already final. Most prior authorization denials are never appealed — not because the denial was correct, but because the patient lacks the knowledge, energy, or representation to challenge it. The program has been operating for months. The accountability mechanism the EFF is attempting to create through litigation would, if it succeeds, arrive after a substantial volume of consequential decisions have already been made with no public review.
The Accuracy Problem and the Accountability Problem Are Not the Same Argument
Practitioner skepticism about AI in clinical settings tends to focus on performance — whether models are reliable enough to use at all. That argument has been circulating in healthcare AI communities for years, and one practitioner who works in AI evaluation for medical research acknowledged the split directly: the same technology produces "extraordinary work and the crazed hallucinations and really bad output" in ways that are difficult to predict in advance. The Royal College of Radiologists published post-deployment monitoring guidance for AI medical imaging devices in 2026 , reflecting growing institutional recognition that deploying AI without structured safety reporting creates exactly the risk environment WISeR now represents.
But the EFF's lawsuit is not primarily about whether WISeR is accurate. It is about whether the public has any way to find out. Those are separable questions, and conflating them obscures the more fundamental problem: a program can be reasonably accurate on average while still denying care to individual beneficiaries in ways that a disclosed methodology would flag as outside the model's intended scope. Without the documentation, no one outside CMS can make that determination — not researchers, not advocates, not the beneficiaries whose care it is.
Who Bears the Cost When the Algorithm Is Wrong
The KFF poll that circulated on Bluesky the same week as the EFF filing added a dimension the transparency debate alone cannot carry: a substantial share of Americans are turning to AI for health information not because they prefer it, but because they cannot afford medical care. The population that poll describes overlaps heavily with the Medicare beneficiaries WISeR is affecting — people for whom an incorrect prior authorization denial is not a bureaucratic inconvenience but a foreclosure on treatment they cannot access any other way.
That overlap is not incidental. It is the structural condition that makes opaque algorithmic care denial especially punishing. The people least positioned to navigate an appeal process are the people most dependent on the initial determination being correct — and the people who, as the KFF data suggests , are already substituting AI for professional care because the professional care is financially out of reach. WISeR is operating precisely at the point where that substitution becomes involuntary.
What the FOIA Process Will Force Into the Open
FOIA litigation is slow and the outcome is uncertain, but the process itself generates accountability independent of the verdict. If the court orders disclosure, CMS must produce documentation of WISeR's design — training data provenance, decision criteria, accuracy benchmarks, and any internal assessments of the program's performance since January 2026. If those documents show that CMS launched the program with strong internal evidence of its reliability, the EFF's campaign will have served to publish what should have been public from the start. If the documents show something else — gaps in validation, known error categories, populations the model performs poorly on — the seniors already denied care under WISeR will have been harmed by a program whose limitations CMS chose not to disclose.
The EFF has already won the more important argument: that a federal AI program adjudicating care for millions of seniors requires public documentation. CMS's refusal to provide it without a lawsuit is the answer to the question the EFF was asking.
The story so far
CMS launched WISeR in January 2026 with no public methodology disclosure; the EFF's FOIA suit has put every prior authorization denial since then on a foundation that CMS cannot defend without releasing documentation it has withheld.
Frequently Asked
- Why did CMS launch WISeR without publishing its methodology or error rate?
- CMS has not explained its rationale publicly, and the EFF's FOIA lawsuit exists precisely because the agency declined to disclose program documentation when asked. Federal agencies are not legally required to publish AI methodology before deploying it — the EFF's argument is that a program making care determinations for Medicare beneficiaries triggers a public accountability obligation that CMS has so far refused to meet. The lawsuit will force that question into court.
- What should Medicare patients do if they received a prior authorization denial since January 2026?
- File a formal appeal immediately. WISeR denials since January 2026 were made by an AI program whose criteria CMS has not disclosed — that undisclosed methodology is now the subject of active litigation, which means the evidentiary basis for those denials is contested. Patients who cannot navigate the appeals process alone should contact their State Health Insurance Assistance Program (SHIP) or a patient advocacy organization. The denial is not final until the appeals process is exhausted.
- What is the strongest argument that WISeR's opacity is acceptable?
- The strongest version holds that prior authorization AI functions like other actuarial tools CMS already uses — statistical models applied to administrative data, not clinical judgment — and that requiring full public disclosure of the model's criteria would allow providers to game the system by structuring requests to match the AI's approval patterns. That argument has real operational logic. It does not, however, address the problem that beneficiaries whose care is denied have no way to evaluate whether the denial was within the model's valid operating range — which is the harm the EFF lawsuit targets.
Continue reading
The AI Doctor Visit You Can't Afford to Skip
A third of Americans now use AI chatbots for health information — and affordability, not curiosity, is what's driving the most consequential users.
similarTwo Healthcare AI Narratives, One News Cycle, No Overlap
Health journalists and Bluesky skeptics are covering AI in healthcare as if they inhabit separate industries — and the gap is now a structural feature of how the technology gets adopted.
similarAI Can Read Your Hospital Bill. It Cannot Fix the System That Created It.
AI tools are finding thousands in fraudulent medical charges while the same conversation ignores that billing complexity is a policy choice, not a technical problem.
Methodology
This story was generated autonomously from 20 source records. An editorial model synthesizes, weights, and cites each source. No human editorial judgment was applied.