outputs without good reason. As noted under the first principle, if the responsible staff member and the organization have rigorously reviewed all expressions and factual content and are able to assume full responsibility for what is published, generative AI can be regarded as merely one tool among others. That said, even with human review, there are cases in which AI’s contribution is indispensable to the core substance of an output. Examples include deriving analytical results through AI tools, or producing creative works such as music, images, or videos using generative AI. In such cases, it is appropriate to disclose whether generative AI was used and how it was used—namely, in what manner generative AI contributed to the output. In particular, for outputs that may be confused with reality, such as deepfakes, disclosure of the use of generative AI is necessary in order to prevent confusion among audiences. How to apply the principle of transparency was one of the most debated issues during the discussions that led to the development of this guide and policy framework. Concerns were raised that the criterion—“where generative AI has played a substantive role in producing an output, or where its use may cause confusion”— is inherently ambiguous, and that this ambiguity could allow organizations to arbitrarily decide not to disclose their use of generative AI. However, it is important to reiterate that the purpose of this guide is not to make legal determinations or to establish rigid, objective standards. Decisions about when and how to disclose the use of generative AI should instead reflect each organization’s ethical standards and the outcomes of its internal deliberations. Generative AI Guide for Civil Society