The European Parliament’s CULT Committee held a stakeholder consultation on September 2 to feed into its own-initiative report on Cultural and creative sectors in the age of AI. The timing couldn’t be sharper: the European Commission is only just starting to draft its long-awaited AI strategy for the cultural and creative sectors, while also reviewing the Digital Single Market (DSM) Copyright Directive and weighing targeted fixes to Europe’s copyright environment. The Parliament’s report is expected to set out its main requests on these issues.
In a compelling submission, FIA has taken a hard look at how EU copyright policy has evolved since the DSM Directive was adopted, at a time when AI training wasn’t even on the radar. Since then, the commercial text-and-data-mining exception under Article 4 has been stretched to cover AI training too: a real break from how copyright had always applied, and one that flips the usual default on its head: protected works and performances can now be used for free unless rightsholders manage to validly reserve their rights and oppose such unlicensed use.
In practice, performers rarely even know when their performances, voice or likeness are being used for training purposes — and when they do find out, exercising a meaningful reservation is next to impossible. The tech industry, for its part, conveniently disregards opt-outs on the grounds that they aren’t expressed through “universally accepted” machine-readable protocols (protocols that, tellingly, never quite materialise) all while declining to engage in good-faith licensing discussions with performers’ representative organisations. We also flagged the abysmal lack of transparency that the EU Code of Practice has done little to fix, and pointed out that there has been almost no appetite at policy level to address content scraped before the Directive even entered into force, and thus when it was plainly illegal to use this content.
Our central request is simple: put consent back at the heart of the system. Performers should have a real, fair opportunity to say yes or no to their work being used for AI training, on top of any use for making digital replicas (deep fakes). And for that choice to mean anything, it needs teeth: effective mechanisms for collective and individual bargaining – shielded from presumptions of transfer and the overwhelming, and often abusive, buyout practices that dominate the sector – plus collective management of rights reservations when licensing directly to users, as opposed to productions where rights are consolidated by a producer entity.
But copyright can’t do this job alone. FIA is calling for a harmonised EU framework protecting voice, image and likeness, given how many gaps the current patchwork of national rules leaves wide open. Such a framework would give every individual in the EU, not just performers, a real, effective means to have unlicensed digital replicas (deep fakes) taken offline, and a solid legal basis to claim damages and compensation.
Voice work offers an early warning sign of what’s at stake: in addition to dramatic jobs losses, cultural diversity itself stands to suffer. Dubbing and voice-over work is a craft, not a corner to be cut: one that carries linguistic nuance and cultural texture into films, series and games across Europe. Safeguarding it as work done by humans is essential if Europe’s audiovisual landscape is to stay genuinely diverse, rather than being reduced to a uniform, AI-generated soundtrack.
With so much of our sector dependent on public subsidy, public funding carries a responsibility to lead by example, not work against its own workforce. We believe it should be made conditional on AI not being used to displace jobs that performers would otherwise do; a principle we’re pressing hard for as AgoraEU, the EU’s new €8.6 billion programme for culture, media and civic values for 2028-2034, is being negotiated between the Parliament and the Council. It’s one of FIA’s key asks for that programme, and we intend to keep pressing it.
