South Africa: POPIA & FAIS/FSCA
za-fsca@1.0.0 · ZA · 3 clauses
Evidence requirements for South African financial-services organizations using AI-generated output: POPIA's automated decision-making safeguards and the FAIS General Code of Conduct's accountability for advice and client communications. Generic and enterprise-wide: any AI artefact with legal or client effect gets a named, organization-attested human reviewer.
EU AI Act: Article 14 human oversight
eu-ai-act@1.0.0 · EU · 2 clauses
Evidence that natural persons effectively oversee high-risk AI systems as required by Regulation (EU) 2024/1689. Each attestation is a per-output record that an identified human understood, reviewed, and where necessary intervened in the system's output.
GDPR: Article 22 automated decisions
gdpr-art22@1.0.0 · EU · 2 clauses
Evidence that decisions producing legal or similarly significant effects are not based solely on automated processing, and that the safeguard of human intervention is real, exercised by an identified member of staff with authority to change the decision.
US: NIST AI Risk Management Framework
nist-ai-rmf@1.0.0 · US · 2 clauses
Alignment evidence for the NIST AI RMF 1.0: defined human oversight roles for AI system output (GOVERN) and demonstrable post-deployment human review of system behaviour (MANAGE). Voluntary framework; attestations give each control a per-output, independently verifiable record.
Published versions are immutable forever; corrections and new clauses ship as new versions published alongside the old ones. See governance for the rules, and conformance for what implementations must uphold.