• One way to ensure transparency of an AI system is to prepare documents of diverse results from the development, test, and distribution processes of the AI model or service. If there are detailed documents of the model’s specification, you may transparently disclose the purpose of the model, input, output, performance, biases, and reliability upon user’s request to provide information about the AI model.
• IBM and WEF present measures to ensure transparency in AI systems by preparing model specification documents. IBM provides examples of documents to explain the key information and components of an AI model as needed without revealing the algorithm of the developed system.