View a markdown version of this page

RAIGT01-BP01 Develop a transparency strategy - Responsible AI Lens

RAIGT01-BP01 Develop a transparency strategy

For each identified stakeholder group, choose a transparency strategy (for example, blog post, user guide, FAQs, system documentation, service card) and identify appropriate distribution channels to provide downstream stakeholders information to make informed decisions about the AI system.

Level of risk exposed if this best practice is not established: High

Implementation considerations

  1. Build a transparency format selection process that matches each stakeholder group's needs and technical capabilities with appropriate communication methods like secure portals, public documentation, or interactive guides. Test different formats with sample stakeholders to see which ones are effective in assisting them to make better decisions about using your AI system.

  2. Identify how to reach each stakeholder group effectively, whether through existing professional networks, reporting systems, or direct user interfaces. Set up pilot channels to test delivery effectiveness before rolling out to stakeholders.

  3. Create content development workflows that produce stakeholder-specific information covering system capabilities, limitations, risks, and decision-making guidance tailored to each group's expertise level. Build templates and review processes to keep information consistent while allowing customization for different audiences.

  4. Build automated update systems that track when your AI system changes and trigger content revisions across different distribution channels to keep stakeholder information current. Set up monitoring to catch when outdated information is still circulating and create processes to quickly correct or redirect stakeholders to updated materials.

  5. Engage your organization's leadership to determine public disclosure requirements by identifying information appropriate for public sharing versus proprietary details restricted to internal audiences. Implement a structured publication review and approval process for AI system documentation. This mechanism safeguards sensitive AI design and performance information while maintaining appropriate transparency to different stakeholder groups.

Resources

Related documents: