2. Key Issues Raised by Roundtable Participants
Madeleine Varner from the newly established FTC Office of Technology moderated the roundtable discussion. Participants from different creative fields shared their thoughts about changes and challenges arising from the use of Gen AI.
Participants generally noted that they are not inherently opposed to technology or Gen AI, but rather are opposed to the current model of Gen AI that uses creative inputs for training without proper protections for human creators. Many participants pointed out that Gen AI tools have been incorporated into their respective fields and improved the creation process. Duncan Crabtree, Chief Negotiator at SAG-AFTRA, stated that he believes Gen AI should augment human creativity rather than replace it, and John K Painting, from the American Federation of Musicians of the United States and Canada, acknowledged that Gen AI will bring significant changes and emphasized the need to face these changes head-on rather than resisting progress.
The panelists shared their views on three broad topics of relevance to antitrust practitioners. First, participants discussed the definition of “art,” and the terms under which Gen AI should compete against humans in the market for creative works. Second, roundtable participants discussed the licensing of content and what benefits human creators should derive from the use of Gen AI. Panelists suggested variations on a “Four Cs” framework to think about this issue: Consent, Control, Credit, and Compensation. Third, participants provided views on the potentially disruptive effects of Gen AI technologies on labor markets, such as downward pressure on wages, unequal bargaining positions in the negotiation process, and insufficient anti-exploitation and anti-discrimination guarantees.
2.1. What is Art?
One of the key questions raised by several participants was: “What is art”? Each panelist that asked this question concluded that the definition of “art” should involve some form of human creativity and intelligence. Some participants expressed a concern that if Gen AI firms can generate content of sufficient quality at a much lower cost, these firms may dominate in creative fields, leaving no room for human artists. However, if consumers maintain a preference for art generated by human artists, then competition between Gen AI and human artists is likely to be less fierce. Participants suggested that regulations and disclosures may help to protect human creators and ensure that a sufficient share of the art market remains available to them even with the rise of AI generated content.
In their discussion, panelists emphasized two potential competitive concerns. First, some participants argued that competition between humans and Gen AI models is inherently unfair, as AI essentially replicates and predicts human forms of creative content. The issue is particularly salient when a Gen AI model is prompted to produce content “in the style” of an artist whose work was used to train it. In this scenario the author must effectively compete with their own work. Second, concerns were raised about the quality of Gen AI works, with fears that Gen AI may create opportunities for low-quality content to crowd out higher-quality content. For instance, Neil Clarke, editor of Clarkesworld Magazine, described receiving thousands of AI-generated stories submitted to his magazine portal for a review, making publication decisions significantly more time-intensive.
These two concerns also encompass issues related to the FTC’s consumer protection mandate and questions of whether AI-generated art should be labeled as such. In her opening remarks, Chair Khan stated that “from electronics to organic eggs, consumers expect to be told the origin of a product and its authenticity. Consumers make choices based on that information. The same will likely be true with AI.” Participants echoed this sentiment, discussing whether regulations should be introduced requiring Gen AI companies to explicitly mark their content as artificially generated so that consumers can easily distinguish it from the content created by human artists.
Participants also discussed how issues relating to Gen AI implicate the FTC’s anti-fraud goals. Tim Friedlander, President and Founder of National Association of Voice Actors, suggested that the ability to create synthetic creative works can be used to deceive customers, which could lead to the spread of false and misleading information. As examples, he pointed to misinformation around elections and the fraudulent use of generated voices by kidnappers to deceive people by simulating the voices of their loved ones to collect ransom. Mr. Friedlander argued that fraudulent voice generation could also implicate those whose voice is being generated without their authorization by making it appear that they endorse positions that they do not actually hold.
2.2. “Four Cs”
The main recurring topic across all participants was the need for regulations that enforce what participants referred to as the “Four Cs”: Consent, Control, Credit, and Compensation.
Consent: Roundtable participants demanded that Gen AI firms obtain explicit consent from creators before incorporating their works in training datasets. This issue is particularly salient amid recent allegations of web-scraping and use of pirated materials like the ‘books3’ database. More than one participant noted that they were named plaintiffs in litigation against Gen AI firms relating to copyrighted works being used to train AI models without the consent of the creator.
An important distinction raised by participants is whether consent should operate on an opt-in model or an opt-out model. The distinction is relevant to Chair Khan’s opening remarks around existing rules applying to new technologies—participants claimed that current copyright law establishes a ‘permission-first’ model and accused Gen AI companies of shifting the burden in a way that violates existing laws. During the Q&A session, participants unanimously agreed on the necessity of opt-in consent, citing the impracticality of opt-out as a viable solution for three reasons. First, the proliferation of Gen AI companies makes tracking all necessary opt-outs impossible for creators. Second, the current lack of transparency around training data sets imposes additional administrative burden on artists to determine if their work has been used in Gen AI model training. Third, while opt-out decisions typically apply to future use of the creator’s content, past use of that content has a potentially permanent imprint on a trained AI model.
Control: Many roundtable participants said that they were not opposed to Gen AI but rather were opposed to the lack of control that artists have over how their work is used. For example, Mr. Friedlander stated that he is developing his own “synthetic voice clone,” which he aims to license to firms for situations where human voice acting is not possible. As long as sufficient protections over control of inputs and outputs are in place, Gen AI was seen by many roundtable participants as a useful tool to support human artists in their creations.
Credit: Roundtable participants argued that many creative fields have strict standards around crediting, and these standards should also apply to companies using Gen AI. However, the roundtable participants did not address how inputted creative works can be credited to the outputs generated.
Compensation: Participants pointed out that in the status quo, Gen AI companies do not pay individual creators to use their works. Steven Zapata, a representative of the Concept Art Association, discussed a compensation scheme in which artists would be paid per generation for using their works and names, including back-pay for content that has already been generated. However, he also alluded to what he called the “Shutterstock model,” in which creatives are paid fractions of pennies for their work. He said that such a compensation model would likely be unsustainable. In terms of the FTC’s mandates, several participants argued that negotiation of terms should be on equal footing with large tech corporations to avoid the entrenchment of dominant positions in the market.
2.3. Effects on Labor Market
Roundtable participants voiced concerns about Gen AI potentially “replacing” their work in fields as diverse as modelling, concept art, and orchestral performances for theater productions. Sophisticated models can generate content on demand in a particular artist’s style, leading to decreased demand for human-generated content. This could lead to a few large Gen AI firms dominating the labor market for creative works. Markets where the introduction of Gen AI could significantly impact content creators’ earnings might attract the FTC’s attention, as suggested in the recent Draft Merger Guidelines, which emphasized scrutiny on monopsonization and dominance in labor markets.
The existence of dominant Gen AI firms can also lead to unequal bargaining power between individual content creators and large technology firms. Participants mentioned two primary mechanisms to mitigate bargaining power disparity: collective bargaining and legislative lobbying. However, many creators, such as models or visual artists, work as individual businesses. This means they cannot form unions and use collective bargaining to enhance their negotiating positions. Participants also suggested that the existence of dominant firms can lead to exploitation or discrimination. For example, Sara Ziff, executive director of the Model Alliance, voiced discrimination concerns by pointing to examples of “digital blackface”—referring to a situation in which companies use an AI to generate models of color to project a false image of diversity.
3. Closing Discussion
The roundtable concluded with FTC Commissioner Alvaro Bedoya reminding the audience of the importance of human input for developing creative work. He noted the radical creativity of Walt Whitman as an example of the type of creative content generation that only human artists could achieve. He argued that art is not about writing a hundred bad copies of Shakespeare but about expanding what is possible in culture, and that Gen AI is incapable of introducing concepts it has never encountered in training. Therefore, he concluded, creative work requires human intelligence as its primary input, so Gen AI will never eliminate people from creative fields. Commissioner Bedoya closed the discussion by expressing his concerns about instances of exploitation known to him and assuring roundtable participants that the FTC is paying close attention to their experiences and to the rapid developments of technology in this space.