The Bundesrechnungshof, together with four other supreme audit institutions (SAIs), has published the first internationally agreed white paper for audits of machine learning and artificial intelligence (AI) algorithms in public administration.

The white paper, prepared jointly with the SAIs of Finland, the Netherlands, Norway and the United Kingdom, addresses the audit of ML algorithms used in public administration. It is based on the participating SAIs' audit experiences with traditional software development projects and IT procedures, pilot projects/audits on ML and AI, and scientific findings. It provides government auditors with a working tool for ML audits in public administration.

The white paper contains an audit catalogue with guidelines, risk-based audit proposals and guidance for auditors. An assistance tool makes it possible to create individual questionnaires that can be adapted to the audit process.

The English white paper and the assistance tool are made available by the Norwegian SAI on a website. The January 2021 issue of the International Journal of Government Auditing contains a summary article on the White Paper.