Data and AI

• Use data and AI only for an agreed purpose and within the approved scope.

• Minimize client and personal information.

• Do not use client material to train a model unless explicitly authorized.

• Make human review and accountability visible for consequential decisions.

• Document material sources, assumptions, limitations and third-party dependencies.

• Use secure transfer and approved tools for sensitive information.

• Escalate incidents, uncertainty and limitations rather than concealing them.

• Review impacts on access, equity, safety and affected stakeholders.