Automated eligibility and fraud decisions face the highest public-law scrutiny of any sector
Benefits triage, fraud-detection scoring and case-prioritisation systems combine UK GDPR ADM obligations with public-law duties around fairness, proportionality and non-discrimination. Public bodies and their outsourced delivery partners need an assessment that stands up to both ICO and judicial scrutiny.
Existing public-sector automation is already under the microscope
The DWP's Universal Credit fraud-detection model — one of 58 automations deployed across the department, 38 currently active, handling over 44 million claims — has faced direct scrutiny over cost-effectiveness and fairness, including its own published fairness assessment acknowledging that non-UK nationals were more likely to be referred by the model, even though rejection rates following human review were found to be equivalent. This level of transparency and challenge is the standard public bodies should expect their automated systems to be held to.
Eligibility-triage, fraud-scoring and case-prioritisation systems that determine who receives a benefit, faces investigation, or is prioritised for a service decision are prime candidates for Article 22A classification, and public bodies face additional layers of accountability through judicial review risk, the public-sector equality duty, and heightened public and parliamentary scrutiny that private-sector deployments rarely face to the same degree.
A local authority using automated fraud-risk scoring for benefit claims
A local authority's fraud-detection tool scores incoming claims and automatically flags a proportion for investigation, delaying payment pending review. No published fairness assessment exists, and the authority cannot currently demonstrate what proportion of flagged claims are ultimately found to be fraudulent. Illustrative example — not a real client or verified statistic.
Flagged claims are delayed automatically with no published fairness data and no clear appeal route.
The scoring system is classified, a fairness/equality assessment is documented, and a clear human-review and appeal pathway is evidenced.
Where to start
- Classify fraud-scoring, eligibility-triage and case-prioritisation tools against Article 22A, with particular attention to disparate impact across protected groups.
- Publish or document a fairness assessment for any system affecting benefit, service or investigation decisions, following the DWP's own precedent.
- Confirm compliance with the public-sector equality duty alongside the standard ADM safeguards.
- Build an accessible, evidenced human-review and appeal route for every automated flag or decision.