Croissant Architecture

Layer 3: Responsible AI (RAI)

The RAI layer addresses the critical need for machine-readable documentation regarding the ethical and social impacts of datasets.

Documentation as Data

Proposes a machine-readable way to capture and share metadata traditionally found in Data Cards. This makes documentation searchable and reusable.

Lifecycle Traceability

Records granular details on how a dataset was created, processed, and enriched. This provides transparency for downstream AI safety assessments.

Social & Ethical Impact

Back to Overview