Skip to content
AID-2024-0004August 2024 (ruling); system launched 2019critical

TennCare Connect / TEDS automated Medicaid terminations

Tennessee's $400M+ automated Medicaid eligibility system wrongfully terminated or denied coverage for tens of thousands, and a federal court ruled the program violated the Medicaid Act, due process, and the ADA.

Wrongful automated decisionHigh-risk approval gateNamed approval gate

What happened

On August 26, 2024, U.S. District Judge Waverly Crenshaw of the Middle District of Tennessee ruled in A.M.C. v. Smith (No. 3:20-cv-00240) that Tennessee's Medicaid program, TennCare, violated the Medicaid Act, the Fourteenth Amendment's Due Process Clause, and the Americans with Disabilities Act. The Tennessee Eligibility Determination System (TEDS), which runs the TennCare Connect portal, was built by Deloitte Consulting and other contractors (after Northrop Grumman was fired) and launched in 2019 at a total cost of more than $400 million. Programming errors and design defects improperly screened out thousands of enrollees. The court found that enrollees were placed in the wrong households or dropped from their correct households, causing renewal notices to be sent to incorrect addresses; that the system failed to consistently load eligibility data such as Social Security receipt, disability status, and marital status; and that termination notices were misleading and deficient, depriving enrollees of fair-hearing rights. In one documented example, a single mother and her five children wrongfully lost coverage. The class action, originally filed in 2020, covers a class of over 100,000 individuals whose benefits were terminated plus over 100,000 qualified individuals with a disability. Plaintiffs were represented by the National Health Law Program, Tennessee Justice Center, National Center for Law and Economic Justice, and Selendy Gay PLLC.

What the agent did

The TEDS automated eligibility system made and executed coverage determinations, terminating and denying Medicaid benefits for enrollees based on flawed data handling and defective logic. State officials administered the program built on this system; the wrongful terminations were produced by the automated determinations rather than by individualized human review, and the deficient notices were system-generated.

The irreversible effect

Tens of thousands of low-income adults, children, and people with disabilities lost or were denied Medicaid health coverage, in some cases for extended periods, without adequate notice or a meaningful opportunity to contest the decision before losing access to care.

Root cause

Programming errors and design defects in the Deloitte-built TEDS system caused enrollees to be placed in the wrong households or dropped from correct ones, failed to consistently load eligibility data (Social Security receipt, disability status, marital status), sent renewal notices to incorrect addresses, and generated misleading or deficient termination notices. The absence of adequate human review and fair-hearing safeguards let erroneous automated determinations take effect.

How a maker-checker control would have refused it

This harm was caused by a government eligibility system executing automated determinations, not by an AI assistant refusing a request, so a maker-checker refusal did not and could not have literally fired here. Framed as a hypothetical: a maker-checker or approval-gate control would treat termination of a person's health coverage as a high-risk, hard-to-reverse action requiring an independent checker to verify the underlying eligibility data and the adequacy of the notice before the termination took effect. Such a gate, applied at scale, would have surfaced the household-assignment errors, missing eligibility data, and deficient notices that the court found, and would have held the terminations pending human review rather than letting the system act on defective determinations. The court's ruling that enrollees were denied fair-hearing rights is essentially a finding that no such independent check existed before the consequential action.

Accuracy and corrections

This entry describes a publicly reported incident and is compiled from the primary sources listed above. Where an account is a legal allegation rather than an established finding, the entry labels it as such. Summaries can still contain errors. If you can document a correction, email hello@makerchecker.ai and we will review and correct it, with the change noted, within 14 days.

See it for yourself

Reading is one thing. Watch it block an agent.

One command starts the demo: an agent stopped from signing off its own work, and the signed evidence file an inspector can check for themselves.

Designed against the rules your auditors already enforce.