Model evaluation amendments and continuing conditions

AIR.01.01officeUnion-wide competence

Purpose and competence

Model evaluation amendments and continuing conditions is maintained for Model evaluation. Its scope follows the office’s mandate: Evaluate a model's supported behaviour against its actual consequential use.

A model's aggregate score conceals a harmful failure for a small resident group

Record specification

Record identifier: AIR.01.01/R3
Responsible custodian: Model evaluation
Source evidence: model version; test coverage; affected decisions
Decision output: AI model evaluation finding
Publication status: Draft record specification / no signed decision
Review route: Union Ombudsman / Union Court of Justice

Custodian and access

Model evaluation assessment lead → · Public records access procedure

A change that may reopen the ai model evaluation finding

A model's aggregate score conceals a harmful failure for a small resident group; examine the supported subgroup evidence.

Instrument affected: AI model evaluation finding
Receiving custodian: Model evaluation assessment lead

Access to this record

Model evaluation assessment lead / contact · Public records procedure · Common custody and accessible-format rules