The organization should implement specific, measurable controls within its AI system development processes to ensure adherence to established responsible AI engineering principles. Regular verification activities should be conducted to confirm that these foundational objectives are consistently achieved throughout the AI system's lifecycle, from initial design through to deployment and maintenance.

The organization should establish and document clear objectives for the responsible development and use of AI systems. These objectives should be aligned with the organization's AI governance framework and overall strategy. When defining the objectives, the organization should consider ethical principles, potential societal impacts, and the expectations of relevant stakeholders. The objectives should also guide the entire AI system lifecycle to ensure AI systems are used in a safe, fair, and transparent manner. Where practicable, the objectives should be measurable.

The organization should define and openly communicate its responsible AI objectives to ensure they are visible to all relevant stakeholders. Objectives should relate to fairness, accountability, explainability, safety, privacy, and accessibility.

The organization should document how it meets these objectives in order for decision-making to be clearly recorded and auditable. This should the rationale for choosing third-party or internal AI solutions. The organization can demonstrate transparency by embedding meaningful human oversight of its AI systems in the following ways:

• Giving human reviewers the authority to review and override AI decisions where necessary
• Disclosing compliance requirements in deployment documentation
• Monitoring system performance for the purposes of demonstrating ongoing accountability
• Maintaining clear reporting channels for concerns about AI outputs and promptly communicating changes in performance to stakeholders
• Ensuring automated decisions documented and subject to human review

The organisation should establish and provide clear guidelines for individuals assigned to oversee AI systems. These guidelines should detail the procedures for monitoring performance, interpreting outputs, and intervening in operations. Instructions should cover how to override or halt the system and include measures to raise awareness of automation bias.