Such disclosure may be appropriate even where the organization has carried out a certain level of review of the content. For example, when generative AI is used for data analysis, it may be difficult for humans to identify all potential errors. In the case of artistic or creative outputs, the absence of any disclosure may lead audiences to assume that the work was created entirely by a human. At present, many generative AI outputs are still somewhat recognizable as such, but as the technology advances, this boundary will become increasingly blurred. In cases such as deepfakes—where images or videos are deliberately manipulated to resemble reality—confusion among audiences may escalate into more serious harms beyond mere misunderstanding. Deepfake technology may be used not only for illegal purposes, such as deepfake sexual abuse material, but also for the creation of lawful works. For example, civil society organizations may use deepfakes in documentaries to protect the identities of LGBTQ+ individuals, or to produce parody works that criticize those in positions of power. In such cases, if disclosure requirements would interfere with the audience’s experience of the work, disclosure may be provided in a manner that does not undermine its enjoyment (for example, by including a notice in the credits). Indicating that a work involves deepfake technology is also a requirement under the EU AI Act, and it is highly likely that similar regulations will be adopted in an increasing number of countries. 90 91