Performers’, authors’ and other creative workers’ organizations at EU level jointly call the EU Commission, Parliament and Council, convening in trilogue for final amendments to the EU AI Act, to ensure that AI training and deployment uphold the highest standards of transparency. Despite positive amendments suggested by the EU Parliament, requiring providers of foundational models to “make publicly available a sufficiently detailed summary of the use of training data protected under copyright law” the draft regulation falls short of empowering artists and creative workers to effectively monitor how their protected works and performances, as well as their personal data, may be used for AI training purposes. This is crucial for them to assert their right to autorise or prohibit such use.
Transparency must extend to all protected content and personal features used for training purposes and these lists should be readily accessible to anyone suspecting use without prior informed consent. Failure to do so may jeopardize their ability to invoke protections under the GDPR or the CDSM directive.
In addition, all use of AI should be openly disclosed to the public, with warnings about the inclusion of “deep fakes” in the content they consume, even in cases where prior consent has been granted. Broad exemptions from labelling obligations when the transformative use of AI is deemed “necessary for the exercise of the right to freedom of expression and the right to freedom of the arts and sciences” risk allowing most deep fakes to go unnoticed. In addition to the important societal implications that this may have, the potential backslash from an uninformed audience onto the artists should not be underestimated. FIA strongly asserts that audio and video labelling obligations are perfectly compatible with, and in no way impeding, these freedoms and that, considering the potential negative impacts of deepfakes when they are not recognized as such, exemptions from labeling requirements may only be envisaged when the transformative use of AI falls within a use that is authorized by law (including in the context of certain copyright exceptions).
FIA urges the EU decision-makers to make sure that the AI Act, which is poised to be a pioneering effort by a major regulator to harness the unfettered use of AI, setting a benchmark for global regulation, is enthused with the highest degree of transparency, safeguarding the rights of artists and other cultural workers and mitigating the (high) risks associated with disinformation to protect our societies.
