This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
Between 2013 and 2015, the Michigan Unemployment Insurance Agency implemented the MiDAS automated system to detect fraud in unemployment claims. The system, which operated with minimal human oversight, incorrectly flagged tens of thousands of individuals for fraud, resulting in severe financial and personal consequences including wage garnishment, bankruptcy, and loss of housing. The state eventually acknowledged the high error rate and reached a $20 million settlement with affected claimants.
Michigan’s MiDAS system falsely accused over 34,000 people of unemployment fraud from 2013 to 2015, which reportedly caused financial ruin for many. The automated system was designed to cut costs, but it adjudicated fraud cases without human oversight. That led to an 85% error rate. Victims faced wage garnishments, some lost homes, and some faced bankruptcy. Despite early warnings, Michigan’s UIA defended MiDAS until lawsuits and federal pressure forced reforms. Legislators have been seeking compensation for those wrongfully accused.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
Delegating by humans of key decisions to AI systems, or AI systems that make decisions that diminish human control and autonomy, potentially leading to humans feeling disempowered, losing the ability to shape a fulfilling life trajectory, or becoming cognitively enfeebled.
Also classified as
The MiDAS system failed to perform reliably, operating with a catastrophic 93 percent error rate due to flawed mathematical assumptions and an inability to verify data context.
The state delegated key decision-making authority regarding fraud determinations and financial seizures entirely to the MiDAS algorithm, removing human review and leaving victims with no meaningful immediate recourse or appeal.
AI system
Due to a decision or action made by an AI system
Unintentional
Due to an unexpected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed
The severe financial and personal harms were caused by the automated decisions of the MiDAS system, which produced unexpected and highly erroneous outcomes while pursuing the goal of fraud detection.
Added categories peak at 3: Substantial (Psychological Harm, Indirect)
The MiDAS incident represents a major failure of an autonomous state-level decision-making system that violated the civil rights of 40,000 citizens. While it caused severe local economic and societal distress, its direct national security implications remain negligible to minor.
The MiDAS system operated independently without human supervision, automatically matching database records, generating fraud determinations, and triggering financial seizures based on pre-programmed algorithmic rules.
High Risk: The system is an AI used by a public authority to evaluate eligibility for public assistance benefits and services, which has significant implications for safety, fundamental rights, and public interests.
The primary goal of the MiDAS system was to detect fraudulent unemployment insurance claims (cheating detection) and automate the evaluation of ongoing eligibility (application evaluation).
The reports provide extensive, consistent, and detailed coverage of the MiDAS system, its deployment, the legal proceedings, the number of affected individuals, the error rate, and the specific harms caused. Multiple reputable sources corroborate the core facts.
None that are plausible; the role of the automated system in causing the harm is clear and legally established.
Technology
Robo-Adjudication Algorithm
System made final fraud decisions without any human oversight or review.
Flawed Earnings Calculation
Divided quarterly income by 13 weeks to falsely assume steady weekly work.
Inability to Assess Intent
Algorithm could not determine if discrepancies were due to simple mistakes.
Data Inputs
One-Sided Data Preference
Programmed to automatically assume the employer data was always correct.
Missing or Corrupt Data
System based some of its fraud findings on missing or corrupt database records.
Complex Government Databases
Engineers misunderstood complex data structures and collections processes.
Human Factors
Engineering Hubris
Engineers overestimated capabilities and ignored complex local workflows.
Automation Bias of Overseers
Human overseers were hesitant to question or override automated results.
Confusing Portal Notifications
Claimants missed pings sent to inactive online accounts instead of mail.
Process and Methods
Elimination of Human Staff
Laid off one-third of the agency workforce, removing human oversight.
Flawed Questionnaire Design
Confusing multiple-choice questions led innocent people to self-incriminate.
Inadequate Appeals Window
Strict 30-day window to appeal was finalized before claimants knew of charges.
Management
Prioritizing Cost Over Accuracy
Focused on reducing staff and cutting budgets instead of system safety.
Ignoring Warning Signs
Management defended system because of the surplus revenue it generated.
Contractor Compliance Focus
Developers followed orders blindly without questioning system validity.
Regulatory Environment
Highest-in-Nation Penalties
State law mandated 400 percent penalties, amplifying the financial harm.
Lack of Independent Vetting
No independent interdisciplinary group vetted technology before purchase.
Deficient Due Process Protections
Absence of legal requirements for pre-deprivation human review.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
