POLICY PAPER VILIFY, RIDICULE, DISINFORM Political communication and media trust

The rise of generative AI audio and their use in Risks of mislabelling manipulated content disinformation campaigns in the context of recent elections as AI-generated in Poland and Slovakia is evidence of the multi-modality of threats posed by generative AI, and has exposed the blind- From the review above, is it important to note that AI-related spots in some platform guidelines on manipulated co. [...] Many of that providing a disclaimer in the text of the initial post is not the goals of disinformation actors may be achieved without su cient as this in ormation is easily lost as the media ( or the aid of sophisticated technologies or techniques through example, the accompanying image) is shared and re-shared simply de-contextualising information or misquoting politi- across the internet. [...] legal obligation will be determined by the new European ‘AI O ce’, also to be established by the EU AI Act, which will be responsible for “drawing up of codes of practice at Union level Emerging legal rules and EU AI Act to facilitate the effective implementation of the obligations regarding the detection and labelling o arti cially generated An expanding range of legal rules apply to the uses a. [...] 84 outputs of generative AI systems – such as the creation of deepfakes – and to the development of generative and ot- As discussed above, the providers of social media and other her kinds of AI systems. [...] A stated purpose of the EU AI Act is labels; whether the label’s author – such as, for example, a to protect the ‘integrity’ of and ‘trust in the information eco- social media company – will infuence citizen’s trust in the la- system’ with several of its rules animated by policymakers’ bel; whether it will be easy to remove labels; and whether the anxieties over ‘new risks of misinformation and ma.