Federal AI Prior Authorization Program Fails 53% of Requests, GAO Finds Procedural Breach
Source
Trump admin using AI to deny medical care for seniors in disastrous experimentCenters for Medicare and Medicaid Services (CMS)
What happened
The Centers for Medicare and Medicaid Services launched the Workflow Integration System for eReview (WISeR) pilot in January 2026, deploying AI and machine learning to automate prior authorization decisions for Medicare services across six states. Reporting by Ars Technica, based on documents obtained through Electronic Frontier Foundation litigation, reveals that one vendor denied more than 53 percent of prior authorization requests, while several vendors failed to meet the program's mandated 72-hour decision window — leaving patients waiting months for care decisions. The Trump admin using AI to deny medical care for seniors in disastrous experiment report documents rushed implementation and confirmed technical failures. In May 2026, the Government Accountability Office determined that CMS did not follow proper administrative procedure in establishing WISeR, creating significant legal exposure for the program and raising accountability questions for any enterprise deploying AI in analogous high-stakes decision contexts.
Why it matters
- ·The GAO's procedural non-compliance finding signals that AI deployment in regulated decision-making contexts is now subject to independent audit and legal challenge. Any organization using AI to make or inform consequential decisions — benefits eligibility, insurance claims, clinical authorization — faces parallel exposure if governance and rulemaking procedures were not followed before deployment.
- ·The 53 percent denial rate and missed 72-hour SLA windows demonstrate that model performance monitoring and operational compliance thresholds must be treated as binding governance controls, not aspirational benchmarks. Compliance teams need documented processes to detect and escalate when AI-driven decisions deviate from regulatory or contractual service standards.
- ·Litigation-driven disclosure — the EFF obtained program documents through legal action — shows that internal AI deployment records, including technical failure logs and implementation timelines, are discoverable. Organizations must assume that documentation gaps in AI deployment programs will surface under adversarial scrutiny, making audit-ready records a legal necessity, not just a governance best practice.
Governance controls affected
What to do now
- ☐Audit any AI system currently used to make or support consequential decisions — benefits denials, claims adjudication, clinical authorization — and confirm that human override and escalation procedures are documented and tested.
- ☐Verify that AI-driven decision workflows have binding SLA or regulatory time-window requirements translated into monitored performance thresholds, with escalation paths when thresholds are missed.
- ☐Review internal records related to AI program implementation for completeness and defensibility, on the assumption that litigation or regulatory inquiry could make those records discoverable.
- ☐Confirm that any government-contracted or publicly funded AI program your organization operates or supports followed required administrative or rulemaking procedures before deployment.
- ☐Assess whether current incident classification criteria would capture a pattern of systematically elevated AI denial rates or missed response windows as a reportable AI incident requiring escalation.
What to watch next
The GAO procedural finding and ongoing EFF litigation are likely to generate additional disclosure demands and potentially congressional oversight hearings focused on AI in federal benefits administration. Compliance teams at healthcare payers and government technology contractors should monitor whether CMS issues corrective guidance, suspends WISeR, or faces judicial action — each outcome would set a precedent for how AI-driven administrative decisions are governed across the public sector. The case also increases the probability that pending state legislation, including provisions in the Colorado AI Act SB205 and similar high-risk AI bills, will tighten human oversight requirements for automated benefits decisions specifically. Enterprise teams should track whether the WISeR findings influence rulemaking timelines or enforcement posture at other federal agencies using AI for adjudicative functions.
Stay ahead of stories like this
Get every US AI governance development like this one, plus the rest of the week's developments. Every Thursday.