EP poll not deepfake Armageddon some media predicted - study

(ANSA) - ROME, OCT 25 - The European elections were not the deepfake Armageddon that some media had predicted.
    Yet, in several parts of Europe, deepfakes of politicians have already become a critical part of the public debate.
    It is a phenomenon that cannot be contained with agreements with political parties and online platforms without the ability of governments to enforce such agreements.
    A phenomenon that above all risks undermining voters' trust in the political system.
    This is the snapshot taken by the University of Texas at Austin on the use of generative artificial intelligence (GenAI) in European politics.
    The report, which examines both the European elections and the elections in France and the United Kingdom, focuses in particular on the issue of deepfakes although, experts write, it is not yet a dominant factor in electoral processes.
    Several prominent politicians, including German Chancellor Olaf Scholz, British Prime Minister Keir Starmer, and Rassemblement National leader Marine Le Pen, have been targets of deepfakes, some of which have more satirical undertones.
    "When the discourse becomes muddled," the report says, "voters begin to question the integrity of not just particular deepfakes but all political campaign communications, a real cause for concern." Non-binding voluntary agreements signed by political parties and tech platforms "have not prevented the sharing of political deepfakes," the experts say.
    The report cites the Atlantic Council's Digital Forensics Lab (DFRLab), which found that the far-right EP caucus Identity and Democracy (ID) used GenAI in its election campaigns, violating ID's code of conduct for the 2024 European Parliament elections.
    Similarly, AlgorithmWatch reported that companies like OpenAI have failed to comply with rules that prevented them from creating realistic images of candidates for elections.
    "These examples illustrate how crucial transparency and clarity are," the experts write.
    "When neither political parties nor platforms label content clearly enough, voters are left with a losing battle because the specter of doubt persists." Voluntary commitments from platforms to provenance, transparency, and disclosure are important, but, the report notes, "the ability of governments to enforce" those agreements, by pressuring platforms to follow through on promises of watermarking, detection, and mitigation, is even more important.
    (ANSA).