Independent algorithmic audit proposals have emerged from the Ada Lovelace Institute, AI Now Institute, Algorithmic Justice League, OpenMined, Mozilla Foundation, and academic researchers (Raji, Buolamwini, Mitchell, Whittaker, and others). Common elements include third-party impact assessments, red-team requirements, disclosure of training data composition, and post-deployment monitoring for disparate impact.
These proposals share a common premise: AI accountability requires inspection by parties other than the developer. They underwrite the audit frameworks emerging in the EU AI Act, NYC Local Law 144 (employment AI bias audits), and CFPB/EEOC guidance on automated decisions.