DPE's Response to the U.S. Copyright Office AI Study's Request for Information
October 30, 2023
Dear Ms. Perlmutter,
On behalf of the Department for Professional Employees, AFL-CIO (DPE), I write in response to the United States Copyright Office’s (USCO) August 30, 2023 notice of inquiry regarding the USCO’s study on artificial intelligence (AI). For the purposes of this filing, I am specifically discussing generative artificial intelligence (GAI) technologies. It is challenging to fully discuss all the ways in which these technologies will impact creative professionals.
By way of introduction, DPE is a coalition of 24 national unions, 12 of which are unions representing creative professionals working in the arts, entertainment, and media industries. These unions’ members work as actors, stagehands, craftspeople, choreographers, dancers, directors, directorial team members, editors, musicians, stunt performers, instrumentalists, writers, singers, stage managers, recording artists, broadcasters, audio engineers, cinematographers, and many other creative positions. They help power a sector of the economy that regularly generates four percent of the United States’ gross domestic product (GDP), creates a positive trade balance, and employs more than five million people.
DPE appreciates USCO’s efforts to examine the impact artificial intelligence has on the copyright framework and enforcement as well as its impact on creators’ rights and a changing business environment. This is an issue that directly impacts the members of DPE’s affiliate unions in the arts, entertainment, and media industries. While not typically the copyright holder, many of these middle-class creators earn collectively bargained pay and contributions to their health insurance and pension plans from the sales and licensing of the content they help create. In 2021, for instance, creative professionals represented by the American Federation of Musicians (AFM), Directors Guild of America (DGA), International Alliance of Theatrical Stage Employees (IATSE), Screen Actors Guild - American Federation of Television and Radio Artists (SAG-AFTRA), and Writers Guild of America, East (WGAE) received over $2.8 billion in residuals. In addition, union creative professionals’ future work opportunities depend on legitimate sales and licensing.
AI is a double-edged sword that is empowering and democratizing storytelling and creation to an unprecedented degree in history. It opens doors to new jobs and new markets to monetize creative works in a beneficial way. In addition, creative professionals are harnessing AI as a tool to increase efficiency and support their work. At the same time, AI poses a significant threat of abuse absent proper safeguards.
AI has the ability to upend the creative industries by disrupting employment opportunities, revenue streams, and an effective ecosystem that is shaped by personal services agreements, collective bargaining agreements, and licensing arrangements. Absent safeguards to ensure consent and compensation for the use of copyrighted works and individual intellectual property rights, and appropriate transparency of training sets, AI will be used as a sophisticated, deceptive tool for content theft, unauthorized digital replication of individual’s voices and likenesses, and cultural misappropriation. Developers will be incentivized to train their AI technologies on the creative works and the creative talents that the members of DPE’s affiliate unions develop, design, and bring to life.
The copyrighted works of expression being ingested into AI models would not exist but for the human likenesses, sounds, labors, and creativity of working people like the members of DPE’s affiliate unions. The founders of this country intended to encourage human creativity, not machinery, when it drafted the copyright clause of the U.S. Constitution. Additionally, the First Amendment was intended to protect human expression from censorship and suppression.
AI is not on the horizon; it is already our reality. We are witness to the hyper-realistic replication of performers’ voices, faces, and individual instrumental sounds, and the automation of sound recordings, compositions, performances, film editing, cinematography, and more through GAI. Animators, directors, musical composers, performers, writers, and countless other craftspeople are seeing their creative works fed into AI models, yielding convincing machine- generated results. If today’s policymakers and government officials fail to secure guardrails to this technological revolution, creative professionals will suffer from their inaction. It is no longer a question of being proactive; the time has passed, and the government is now compelled to respond defensively.
DPE appreciates your continued attention to this pressing issue. In the digital era, copyright protections, and other intellectual property rights, have become increasingly vital for union creative professionals who rely on them to safeguard their livelihood and careers. In the following responses, I address questions related to copyright protections. Please note that while this filing concentrates on copyright issues, individual members of DPE are also submitting comments that delve into their specific requirements and concerns, such as international author rights for directors and writers; rights to one’s name, image, voice, and likeness; and related rights for other craftspersons.
As described above, GAI systems have the ability to produce material that would be copyrightable if it were created by a human author. What are your views on the potential benefits and risks of this technology? How is the use of this technology currently affecting or likely to affect creators, copyright owners, technology developers, researchers, and the public?
Modern technologies can aid creative processes and expand the limits of human expression. Among these, AI emerges as a powerful force, capable of pushing frontiers of visual and auditory depictions. Moreover, AI excels in streamlining and automating certain aspects of content creation to the benefit of workers.
Creative professionals have been using several types of AI technologies for decades across diverse mediums, such as music, motion pictures, live performances, and literature. However, it remains crucial for policymakers to acknowledge that the essence of great art, entertainment, and media derives from human emotions and lived experiences, elements that AI lacks. In essence, without people, GAI is culturally without value.
AI models pose challenges for creative professionals and copyright holders. The AI models rely on ingesting copyrighted works of expression and likenesses to operate. AI has the power to mutilate an artist’s creative vision or point of view on any given topic. A machine cannot produce a painting without first ingesting the works of painters, develop an unauthorized television episode without first ingesting episodes from the rest of the underlying series, or produce a portrait without first ingesting the likenesses of others. In fact, ingesting GAI produced works appears to produce inferior, lower-quality results when compared to machine-created content created by ingesting original works of human authorship.
Creative professionals are actively working to adapt to the changes brought by AI, which may necessitate training and upskilling opportunities to stay relevant in evolving industries. Additionally, new job classifications may lend themselves to labor organizing.
Does the increasing use or distribution of AI-generated material raise any unique issues for your sector or industry as compared to other copyright stakeholders?
AI’s threat to jobs and livelihoods has been a serious issue in the entertainment industry. For example, films can be mutilated into virtual reality experiences, or a third party can change the artist’s creative vision and point of view. ChatGPT can regurgitate a screenplay, which has the potential to take the job of a human screenwriter. Voice cloning, as another example, is enabling employers to do away with foreign language dubbing jobs for working performers. Deepfakes will forever seed doubt in what we hear and see and cause serious reputational or emotional harm to those depicted and those who created the underlying work. For these reasons, we need international authors’ rights, likeness and voice rights, and protections for other creative professionals.
Looking at the copyright framework, union creative professionals often have a unique relationship to employers and their intellectual property. Members of DPE’s affiliate unions often work for a variety of production companies over the course of their career (or even in any given month or year), and the copyright for the productions that they create are owned by the employers or contracting party.
Nevertheless, the safeguarding of the rights and interests of these individuals is many times achieved through collective bargaining agreements (CBA). These CBAs play a pivotal role in ensuring that professionals in the art, entertainment, and media industries receive their fair share of the revenue generated by their work. Additionally, union creative professionals may have individually negotiated agreements that go above the terms of their union contract to also receive additional profit participation and/or creative control of their work and likeness.
Are there any statutory or regulatory approaches that have been adopted or are under consideration in other countries that relate to copyright and AI that should be considered or avoided in the United States? How important a factor is international consistency in this area across borders?
The arts, entertainment, and media industries have evolved into global enterprises that cater to audiences worldwide. Creative professionals should benefit from the worldwide distribution of their work. Works created by these artists must be equally protected in the United States and abroad for the system to work.
GAI knows no borders; it is essential for the United States to work with foreign countries and international organizations to develop basic floors and fundamental standards for the ethical and acceptable uses of this technology. We strongly support the commitment of the G7 countries - to which the United States belongs - to respecting material protected by intellectual property rights, including copyright-protected content, and ensuring transparency of data sets.
DPE feels strongly that the United States should not adopt special fair use standards for this technology. It is concerning that certain countries, such as Japan, Singapore, the United Kingdom, and the European Union, have, to varying degrees, elected to pass legislation to grant technology companies permission to ingest copyrighted works of expression without consent or compensation. Such exemptions lack merit and run counter to treaty obligations. For instance, Japan allows for the ingestion of copyrighted works for use in commercial works and does not require a party to first have legal access to those works. The U.K. has taken a more restrained, still problematic, approach by granting an exemption for non-commercial research purposes and the country does at least require legal access. To be clear, DPE does not support any changes that would expand the existing fair use provisions of the U.S. Copyright Act.
This situation may present an opportunity for the international community to strengthen the enforcement of intellectual property laws on a global scale. Just as foreign holders of copyrights encounter obstacles and exorbitant fees to enforce their rights here, American copyright holders face similar challenges when attempting to protect their rights abroad, even when laws allow for extraterritorial enforcement.
Should copyright owners have to affirmatively consent (opt in) to the use of their works for training materials, or should they be provided with the means to object (opt out)?
The United States, or any foreign country, should not adopt an opt-out system. Such a system places the onus on copyright holders and will leave them vulnerable to exploitation by technology companies. Consequently, it would put the members of DPE’s affiliate unions at even greater disadvantage because they would have no enforcement rights. If third parties wish to use copyrighted works to train AI models and generate GAI materials, it should be incumbent upon them to receive permission from the copyright holder.
It is essential that policymakers hold AI companies and users of AI models accountable for the unauthorized use of copyrighted works. In the absence of a legally recognized exemption, ingestion is a violation of the copyright holder’s exclusive right to reproduce or to make a copy. This violation occurs without the need for further distribution, performance, or the creation of a new work. In fact, a pure copy is less likely, not more, to qualify for the fair use defense as there is no new meaning, purpose, or transformation of the original work. It is merely taking something one would pay for in the market.
It is evident that the copyrighted works in questions are of enormous value to AI companies and that value should result in fair compensation for creative professionals. An opt-out system would create an intellectual property framework that runs counter to the principles of copyright law, sanctioning a windfall of riches for technology companies at the expense of copyright holders and hardworking creative professionals. Such an approach would be regressive and would create more problems than good.
Should Congress consider establishing a compulsory licensing regime?
The United States should firmly resist the urge to adopt a compulsory licensing regime. Mandatory licensing schemes devalue works of expression and force copyright holders to unwillingly go into business with third parties.
It is imperative that Congress preserve the open marketplace for licensing and intellectual property transactions, as well as to reinforce the bargaining power of labor unions in securing fair compensation for the use of works in secondary markets. DPE is concerned that lawmakers and regulators may change the rules as a reaction to the volume of copyrighted works required for AI ingestion and the speed at which AI models can produce new works.
The compulsory licensing scheme established for musical compositions, and in some cases, cable television transmission, has often been criticized for devaluing the true market value of a work and denying a copyright holder the fundamental right to withhold consent. It would be problematic to insert an independent collective management organization into the ecosystem, especially if it involves government interference.
In the United States, labor unions are the most qualified entities to oversee collective licensing schemes for their members. Labor unions possess the expertise and capacity to negotiate with employers and establish formulas, especially when dealing with large-scale licensing. For example, through collective bargaining, union members are paid residuals for the reuse of works in secondary markets, such as paid television.
The film and television industry serves as a prime example of a sector that relies on and flourishes from free market negotiation. This applies whether it be the licensing of a single highly valuable motion picture or the right to distribute an entire catalog of works with varying levels of market value.
In order to allow copyright owners to determine whether their works have been used, should developers of AI models be required to collect, retain, and disclose records regarding the materials used to train their models? Should creators of training datasets have a similar obligation?
Yes, to both questions. As seen in the courts, authors are encountering challenges to prove that their work was ingested into the system. For example, Sarah Silverman is having to go to great lengths to show that her book was ingested into the ChatGPT system after it was able to produce a detailed summary of the book.
Without appropriate transparency and robust recordkeeping, creative professionals and the entities that employ them lack the requisite knowledge to enforce their rights.
There are challenges in relying on watermarks and meta data to track files floating around on the Internet or being ingested into GAI models, as bad actors may simply use technology to remove digital fingerprints.
In a closed model, such as those utilized by the major motion picture studios, the companies can collect specific information to better trace the use of copyrighted work. For instance, in the case of the new 2023 Indiana Jones and the Dial of Destiny film, Disney used its own footage of Harrison Ford from previous movies.
DPE’s affiliate unions are, or will be, negotiating AI-specific terms into their collective bargaining agreements. Certain aspects of the use of AI significantly impact the wages and working conditions of labor union members and may constitute mandatory subjects of bargaining. Labor unions would greatly benefit from having access to searchable information about which works are being ingested and/or used to create new works.
Furthermore, if a harmed party or labor union can demonstrate that a work was ingested without a corresponding record, the courts should be prepared to provide relief in the form of monetary damages, costs, and fees. This will ensure that rights holders and labor unions have the necessary tools and incentives to protect their interests in a changing landscape.
Under copyright law, are there circumstances when a human using a generative AI system should be considered the “author” of material produced by the system? If so, what factors are relevant to that determination? For example, is selecting what material an AI model is trained on and/or providing an iterative series of text commands or prompts sufficient to claim authorship of the resulting output?
This is a complex question as the use of GAI in and of itself should not disqualify a work from copyright protection. In fact, this technology is already being ethically used in the creative process as a tool. Judges and policymakers should exercise caution in responding to wholly GAI material or human-prompted GAI material versus creative professionals that use, mold, and direct GAI at a more intricate level where a person or multiple people are providing enough creative control and modification to transform the GAI material into the work of those people.
GAI lacks the capacity to understand the lived experiences of individuals from diverse backgrounds, such as a Latina, a member of the LGBTQ+ community, an immigrant, a soldier, and/or someone who grew up in a specific location. Only when GAI is molded by a person can it serve as a human expression deserving of legal protection. Ultimately, it is only a human who requires the incentive to create new works and deserves to be rewarded for their creative efforts and aspirations. These principles align with the foundational constitutional justifications for copyright law in the first place.
In the context of AI, the law should uphold these principles by distinguishing between works that are by and large machine-generated and those that are the result of human guidance, molding, and creative control. AI-generated materials - including visual, audio, and literary works - lack sufficient human control and should not be copyrighted. A machine-generated prompt is at most an idea, and ideas are not copyrightable.
In closing, thank you again for your attention to this important issue. Absent smart policymaking and requisite safeguards, AI has the potential to upend the livelihoods and economic security of union creative professionals who rely more than ever on adequate and effective copyright protections to earn compensation and benefits in today’s digital era.
If you have any questions, please contact me or DPE’s Assistant to the President/Legislative Director, Michael Wasser at mwasser@dpeaflcio.org.
Sincerely,
Jennifer Dorning, President