③ In the case of generative AI systems that directly interact with external users—such as chatbots or real-time interpretation tools—users shall be clearly informed that they are interacting with an AI system. ④ This organization’s generative AI policy shall be made publicly available, for example through the organization’s website. As explained in the section on principles above, the principles of accountability and transparency remain critically important in the use of generative AI. However, it is neither realistic nor particularly meaningful to label every output that has involved even minimal use of generative AI. This, of course, presupposes that AI-generated outputs have been rigorously reviewed under the organization’s responsibility. If, for example, a report is produced using generative AI and then released externally without any verification of factual accuracy or assessment of potential bias, the report may contain incorrect or biased information. If the organization fails to disclose the use of generative AI in such a case, audiences are likely to treat all of the report’s contents as factual. Should errors later come to light, the organization’s credibility could be seriously undermined. Conversely, if inaccuracies or biases remain undiscovered, false or distorted information may spread further, and the organization cannot evade responsibility for the resulting harm. Accordingly, organizations should make every effort, as a matter of accountability, to verify the accuracy of information and to assess the risk of bias. Where it is difficult to provide such assurances, it is advisable at a minimum to inform audiences that the output was produced using generative AI and that some of its content may contain errors. Generative AI Guide for Civil Society