‘We’re laundering plagiarism’: Music experts speak on AI and copyright issues following Senate forum

WASHINGTON (TND) — CEOs, labor activists, journalists and media industry insiders gathered in Washington, D.C. last week to discuss the challenges to creative industries with the rise of artificial intelligence with members of the Senate.

Particularly, those in the creative industries sought to discuss issues of transparency of AI models and what the evolving systems of artificially generated content mean for copyright and intellectual property.

As conservatives have been concerned about how text-based AI systems like Open AI’s ChatGPT and Google’s Bard learn based on the content fed to them – primarily ‘left-leaning’ journalism in the case of the former – musicians are questioning what sources generative AI programs are learning from in order to develop. For example, if a song is totally generated by an AI program, artists want to know if it was trained using the music of The Beatles or Michael Jackson or Cher – and what does the use of that catalog as a modeling basis mean for the copyright holders of that music as well as the generated music.

Rick Beato, a YouTube personality and record producer known for his work with Shinedown and Needtobreathe, proposed licensing policy for this end, parallel to licenses needed to use music in a film or in public places.

“A music dataset license will tell people what models the music you’re listening too were trained on,” Beato explained in a video on his YouTube channel posted following the hearing.

For example, if they were trained on the Beatles’ music or Led Zeppelin’s music or Stevie Wonder’s music, the consumer needs to know as do the artists and the copyright holders so they can be compensated fairly.

This viewpoint seemed to be shared and adopted by the hearing’s organizer, Senate Majority Leader Chuck Schumer, D-N.Y., who spoke to reporters on Capitol Hill following the event and shared his key takeaways.

“We have some things I think there were a pretty good consensus on,” Schumer said.

First, when it comes to copyright and AI, we have to create and enforce protections for creators to maintain their identities in the age of AI and that may include ensuring some transparency in the AI systems to see what data they are trained on.

Sen. Todd Young, R-Ind., one of the four Senators on Schumer’s bipartisan group in charge of the hearings, elaborated that while there were diverging views in the chamber, the discussions suggested a “balanced” approach would be needed going forward.

“There were some people who had divergent views, frankly, about how to protect the intellectual property of our producers, from artists to other individuals,” Young explained following Schumer’s comments to reporters.

Some favored a more regulatory approach while others said leave it up to the markets and follow existing court standards, court precedents to applying such doctrines as fair use.

Beato may have also been one of those with “divergent views” as he said in his YouTube video that he proposed that all AI-generated music be prevented from being held under copyright (a statement he further said was tacitly supported by Senators in the hearing).

His recap of his testimony was praised by music critic and journalist Anthony Fantano – “the internet’s busiest music nerd,” self-proclaimed – who echoed Beato’s concerns about transparency and intellectual property.

“As we have been seeing with creators from these AI models and looking at the AI art itself, much of this stuff is just a copy or a rehash or a copy of something else,” Fantanto noted in a YouTube video (also referring to AI art here beyond music).

While there is certainly something being generated through this process, what really a lot of these AI models are doing its really just repurposing artistic information it’s already fed. In a way, we’re kind of laundering plagiarism.

“These AI models are just ripping off other things in order to write or create the thing it’s being instructed to,” he added. “And, as it is said in Rick’s video, there has to be some kind of accountability when it comes to what information or art is being but into these AI models in order to create what it is creating.”

While not a part of the hearing, Fantano made his own policy proposal for such considerations around artificial intelligence. Namely, that anything generated by AI needs to be labeled as such formally and officially. He noted how media enterprises like Sports Illustrated have already been caught trying to pass off AI-generated content as organically created, even going so far as to create AI generated headshots for the fictious writers of said articles.

Young thought that the Senate heard some “thoughtful ideas” on how to create that balanced approach to AI regulation in the Nov. 29 hearing. “We can’t get into particulars because, frankly, I’m going to have to go back and unpack some of my voluminous notes, but I thought it was a very helpful session.”