Role of data sheets in AI governance?

• A. They provide detailed descriptions of the dataset used to train an ML model, including creation process, composition, biases, and ethical considerations
• B. They describe licensing terms for the model
• C. They specify hardware requirements
• D. They are optional project artifacts

Answer: (A)

Data sheets for datasets are governance records that document the dataset used to train an AI model. They detail how the data was collected and curated, what it contains, how it was labeled or preprocessed, and any biases or limitations. They also discuss privacy, consent, and ethical considerations, as well as the intended use and potential risks or harms. This level of documentation supports accountability and risk assessment, enabling auditors and stakeholders to understand data provenance, quality, and fairness implications. It also aids reproducibility and regulatory or policy compliance. Licensing terms for the model or hardware requirements belong to separate aspects of governance or deployment, and while some artifacts may be produced in a governance project, robust practice treats dataset documentation as essential rather than optional.