KPMG Trusted AI framework

AI systems should perform consistently and at the desired state of precision within their intended scope.

AI systems should be managed across the AI lifecycle in a manner that safeguards against bad actors, misinformation or adverse events.

AI systems should be managed across the AI lifecycle in a manner that safeguards against harm to people, business, property and societal well-being.

AI systems should be managed across the AI lifecycle to comply with applicable privacy and data protection legal requirements.

The use of AI systems across the AI lifecycle should be warranted by their purpose, and they should be designed to optimise energy efficiency and minimise carbon emissions.

AI systems should provide an understandable explanation for their output, actions, or decisions across the AI lifecycle, with that explanation being appropriate to the use case, context and audience.

Data used in AI systems across the AI lifecycle should be acquired and handled in compliance with contractual obligations, applicable laws, and regulations, and assessed for accuracy, integrity, completeness, relevance and quality.

Across the AI lifecycle, AI systems should include reasonable disclosures so stakeholders are informed when they are interacting with AI, content was AI-generated, or a decision was influenced by AI.

AI systems should be managed across the AI lifecycle to mitigate bias and promote inclusive engagement for all people.

AI systems should be governed across the AI lifecycle with appropriate levels of human oversight, upholding standards of professional responsibility and quality.