International Voice Actor Groups Tell the EU How To Fix AI Laws

United Voice Artists (UVA) and the National Association of Voice Actors (NAVA) have submitted a European Union Artificial Intelligence Act (AI Act) amendment proposal, “seeking to safeguard the rights and interests of voice actors in the rapidly evolving landscape of artificial intelligence and voice technology,” the two global coalitions advocating and promoting the advancement of the voice acting industry announced in a joint press release on August 2, 2023.

On Saturday, July 22, at San Diego Comic-Con 2023, NAVA, with the support of actors’ union SAG-AFTRA, held the “AI in Entertainment: The Performers’ Perspective” panel with Linsay Rousseau (Transformers: War for Cybertron), Ashly Burch (Mythic Quest), Cissy Jones (The Owl House), Zeke Alton (The Callisto Protocol, SAG-AFTRA negotiating member), Tim Friedlander (Record of Ragnarok, president/founder of NAVA), and Duncan Crabtree-Ireland (SAG-AFTRA National Executive Director & Chief Negotiator), exploring how entertainers are adapting to AI, a groundbreaking technology that can not only replicate their voices and movements but also feed machine learning to create entirely new characters.

For example, in early July, GamesRadar reported on Skyrim Mods that used AI to create pornographic material with the voices of video game characters without the actors’ consent, a practice that’s reportedly not banned. During the SDCC panel, the voice actors discussed how existing legal structures can be used to protect individual rights to performance and publicity, which are violated when generative AI uses an actor’s voice without their consent (e.g. the Skyrim controversy).

See the X post alerting the public to the Skyrim controversy below:

“There’s a philosophical difference between how the creators feel and how the employers feel. [They] don’t want to get caught at a competitive disadvantage. But they’re bargaining for a competitive disadvantage and a financial advantage. We’re bargaining for our very livelihoods. You can die on this hill, or you can be killed there,” Crabtree-Ireland said at SDCC, as reported by Comics Beat. “No one should think that voice actors and others are not working together.”

UVA, global coalition of 35 European voice acting guilds, associations, and unions in France, Spain, Italy, Germany, Austria, Belgium, Poland, Netherlands, has united with organizations in Switzerland, the United States, Turkey, Canada, Africa, Asia, and Latin America, representing more than 20,500 artists, to ensure the creative, media, and entertainment industries’ use of AI in the dubbing and voice-over industry “does not harm artistic heritage and human creativity” and preserves the rights of the artists.

The UVA position paper on the EU’s AI Act notes that although the first wave of generative AI focused on images, “more and more generative AI applications are developing in the field of audio and video content,” advances that “are not sufficiently addressed in the current draft of the AI Act.”

The press release states,

The UVA acknowledges the efforts of the European legislators in designing the AI Act as the first European legislation to regulate certain use cases of AI. However, to ensure that the AI Act does not lead to the extinction of human creativity in the entertainment industry through synthetic and cloning AI techniques and the exploitation of artists and their work by infringing their rights, both from a privacy/publicity data protection and an intellectual property rights protection perspective, we have identified the following key areas that should be urgently addressed in the upcoming trialogue negotiations:

• Prohibition to create or expand audio/vocal databases
• Extension of transparency requirements
• Introduction of unfair contract practices unilaterally imposed by generative AI providers

United Voice Artists and The National Association of Voice Actors call upon policymakers and stakeholders within the European Union to carefully consider the proposed amendment, recognizing its potential to elevate the voice acting community while setting a global standard for AI legislation that respects artistic integrity and the rights of creators.

The UVA argues that “addressing the critical issues above will contribute to preserving artists’ rights and preventing the potential harm caused by the illicit exploitation of peoples’ voices as an integral part of their identity, unlawful use of personal data and misleading output by generative AI systems.” With this in mind, the proposed amendments to the AI Act specifically outline what UVA believes to be unfair contract practice regarding the use of generative AI. 

UVA proposed the following provisions to Amendment 398, Article 28a (new) in the EU Parliament Position on AI:

Unfair contractual terms unilaterally imposed by generative AI providers

1. A contractual term which has been unilaterally imposed by the provider
   on users of generative AI or data owners or providers for generative AI
   training shall not be binding on the latter if it is unfair.
2. A contractual term is not to be considered unfair where it arises from
   applicable Union law.
3. A contractual term is unfair if it is of such a nature that it objectively
   impairs the ability of the party upon whom the term has been unilaterally
   imposed to protect its legitimate commercial interest in the input or its
   use grossly deviates from good commercial practice, is contrary to good
   faith and fair dealing or creates a significant imbalance between the
   rights and the obligations of the parties in the contract.
4. A contractual term is unfair for the purposes of this Article if its object
   or effect is to:
   (a) granting unlimited usage rights to input or output data without
   clear restrictions on the scope, purposes, and duration of usage.
   (b) ambiguous or misleading attribution practices for AI-generated
   content.
   (c) granting of exclusive rights to any input into the generative AI
   system without adequate compensation provided in return.
   (d) allowing the creation of fully or partially synthetic personas or
   any part of personas, such as face, voice, expressions, without the
   explicit informed consent of the concerned personas.
5. A contractual term shall be considered to be unilaterally imposed within
   the meaning of this Article if it has been supplied by one contracting
   party and the other contracting party has not been able to influence its
   content despite an attempt to negotiate it. The contracting party that
   supplied a contractual term shall bear the burden of proving that that
   term has not been unilaterally imposed.
6. Where the unfair contractual term is severable from the remaining
   terms of the contract, those remaining terms shall remain binding. The
   party that supplied the contested term shall not argue that the term is an
   unfair term.
7. This Article shall apply to all new contracts entered into force after …
   [date of entry into force of this Regulation]. Businesses shall review
   existing contractual obligations that are subject to this Regulation by
   …[one year after the date of entry into force of this Regulation].
8. Given the rapidity in which innovations occur in the markets, the list of
   unfair contractual terms within this Article shall be reviewed regularly by
   the Commission and be updated to new business practices if necessary.

Overall, the amendment’s provisions are aimed at making sure that when AI is used, the people whose work it learns from receive proper credit and compensation. “There’s no reason that we can’t have some common-sense guidelines around it,” Crabtree-Ireland said at SDCC. “The truth is, in this industry, a handful of huge mega corporations are going to make decisions about how this is implemented. Those executives and people can make good choices, or they make bad choices. We think all of us together can convince them to make good choices.”

(featured image: Striking Distance Studios)