As reported by Music Business Worldwide, the newly proposed Transparency and Responsibility for Artificial Intelligence Networks (TRAIN) Act, spearheaded by Senator Peter Welch, a Democrat from Vermont, aims to navigate the murky waters surrounding the use of copyrighted materials in training generative AI tools. This legislation is specifically designed to empower rights holders who suspect their creative works have been utilized without permission in AI development.
Under the TRAIN Act, rightsholders can approach any U.S. district court clerk to request an administrative subpoena that would require AI developers to produce documentation deemed “sufficient to identify with certainty” whether their copyrighted works were incorporated in the AI training process. This subpoena will only be granted if the rightsholder asserts they possess “a good faith belief” that their work was indeed used to train a specific AI model.
Senator Welch articulated the urgency of this initiative in a statement issued on November 25th, emphasizing the need to address what he terms the “black box” issue associated with AI developers’ reliance on copyrighted material.
He pointed out the opacity surrounding the data sets utilized for AI training, stating, “This is simple: if your work is used to train AI, there should be a way for you, the copyright holder, to determine that it’s been used by a training model, and you should get compensated if it was.”
The act seeks to relieve the burden from rights holders who currently must prove unauthorized use of their materials in legal disputes against AI developers. Recent high-profile cases involving chatbot creator Anthropic, as well as generative AI music applications like Suno and Udio, have underscored the challenges faced by music companies trying to demonstrate the similarity between AI-generated outputs and their copyrighted content. The TRAIN Act intends to streamline this process significantly.
Further emphasizing the intent behind the legislation, Senator Welch remarked, “We need to give America’s musicians, artists, and creators a tool to find out when AI companies are using their work to train models without artists’ permission. As AI evolves and gets more embedded into our daily lives, we need to set a higher standard for transparency.” However, it is important to note that the TRAIN Act does not mandate AI developers to financially compensate copyright holders for utilizing their works; it solely requires them to disclose such usage.
In tandem with the TRAIN Act, another piece of legislation known as the Content Origin Protection and Integrity from Edited and Deepfaked Media Act (COPIED Act) was introduced in the Senate earlier this year. One of its major provisions would render it unlawful for AI developers to leverage copyrighted works for training purposes without obtaining the necessary permissions — a demand that has been strongly advocated by various organizations within the music industry.
In ongoing copyright lawsuits, AI firms have contended that incorporating copyrighted material for training purposes falls under the “fair use” exemption of copyright law. Rights holders challenging these AI companies vehemently dispute this claim, asserting that the concept of fair use was never intended to apply to scenarios involving the massive collection of copyrighted works, which AI tools then use to generate similar or identical outputs. As courts weigh these arguments, critical decisions regarding the fate of copyrighted works in AI training procedures loom ahead.
The TRAIN Act is part of a larger batch of legislative efforts emerging from Congress aimed at regulating AI use. Among these initiatives is the NO FAKES Act, introduced this year, which would permit individuals to file lawsuits if their likeness or voice is misappropriated for AI deepfakes.
Another related bill, the No AI FRAUD Act, has also been proposed in the House of Representatives.
In its pursuit of transparency, the TRAIN Act shares similarities with the Generative AI Copyright Disclosure Act, which Representative Adam Schiff, a California Democrat, initiated earlier this year in the House. Unlike the TRAIN Act, however, the Generative AI Copyright Disclosure Act would necessitate AI companies to send a notification to the Register of Copyrights, outlining “a sufficiently detailed summary of any copyrighted works used.” Tech firms reacted negatively to this proposal, arguing that the sheer volume of materials utilized in AI training would render such notifications impractical.
By contrast, the TRAIN Act alleviates this concern by mandating AI developers to disclose only the use, or non-use, of specific copyrighted works upon request from the rightsholder. This structured approach seeks to enhance accountability while simplifying the inquiry process for rightsholders. As David Israelite, President and CEO of the National Music Publishers’ Association (NMPA), stated, the act “necessitates precise record-keeping standards from AI developers and gives rightsholders the ability to see whether their copyrighted works have been used without authorization.” He further affirmed, “We strongly support the bill which prioritizes creators who continue to be exploited by unjust AI practices.”
Broad support for the TRAIN Act comes from a diverse array of industry groups, including the American Association of Independent Music (A2IM), the American Federation of Musicians, the American Society of Composers, Authors, and Publishers (ASCAP), BMI, Global Music Rights, the Recording Academy, the Recording Industry Association of America (RIAA), SESAC, and Sound Exchange. Additionally, numerous unions and industry organizations spanning film, television, news media, and book publishing have rallied behind the bill.
Mitch Glazier, Chairman and CEO of the RIAA, declared, “Senator Welch’s carefully calibrated bill will bring much needed transparency to AI, ensuring artists and rightsholders have fair access to the courts when their work is copied for training without authorization or consent. RIAA applauds Senator Welch’s leadership and urges the Senate to enact this important, narrow measure into law.” Elizabeth Matthews, CEO of ASCAP, echoed this sentiment, stating, “The future of America’s vibrant creative economy depends upon laws that protect the rights of human creators.”
Moreover, she emphasized, “By requiring transparency about when and how copyrighted works are used to train generative AI models, the TRAIN Act paves the way for creators to be fairly compensated for the use of their work. On behalf of ASCAP’s more than one million songwriters, composer and music publisher members, we applaud Senator Welch for his leadership.” Adding to the chorus of support, Mike O’Neill, President & CEO of BMI, asserted, “Some AI companies are using creators’ copyrighted works without their permission or compensation to ‘train’ their systems, but there is currently no way for creators to confirm that use or require companies to disclose it. The TRAIN Act will provide a legal avenue for music creators to compel these companies to disclose those actions, which will be a step in the right direction towards greater transparency and accountability.”
Σχόλια