RIAA CEO: The Most Dangerous Flaw of AI Music Is That It Eliminates the Economic Incentive for Human Creators to Make Art

Mitch Glazier is the CEO and Chairman of the RIAA. Over two decades ago, the RIAA became a widely hated villain on the internet for suing college students who downloaded music illegally. Now, he sat down with The Verge's Decoder podcast host Nilay Patel to discuss AI music. By the end of the conversation, the two landed on a counterintuitive situation: fans, creators, and record labels are unusually aligned on one side, facing the big tech firms that train generative models and produce AI-generated music on the other.

During the Napster era, the RIAA's opponents were music fans. Mitch was once sent to college auditoriums to convince students not to use P2P file-sharing services. He remembers walking on stage right after Chuck D, and knowing he was about to get booed, he simply pulled a garbage bag over his head and walked out. Nilay had an even more winding personal history with the issue: when he was in college, Napster was banned by his school, which pushed him to become a copyright lawyer. He later represented Kazaa users sued by the RIAA, before realizing he was not cut out for the job and switching careers to become a journalist. The conversation opened with this old shared grudge, but quickly shifted focus to AI.

Mitch stated that the very purpose of copyright is to ensure more art can be created in the future. Artists need to make a living from their work, and if they cannot do that, they will not share their creations with the public. He described AI-generated music as content that "takes two seconds to make, has no lived experience behind it, and in my view, has no soul". The real core problem is what will happen to our culture when machine-made music dilutes or drowns out the economic incentives for human creators to make new work. This argument carries emotional weight, but the legal battle they are fighting is based on a much firmer point: substitution. The conversation repeatedly circled back to the fourth fair use factor, specifically whether AI music harms the potential market for original copyrighted works. The core of the RIAA's lawsuit against Suno is that AI music and human-made music are competing on the same platforms, for the exact same subscription fee, making them direct substitutes.

The key difference between the Napster era and the current AI era is that the RIAA is no longer standing against fans. Mitch noted that back during Napster, fans were on one side, creators were isolated, and companies were on the opposite side; now, AI companies are the isolated party, and everyone else is aligned on the other side. Nilay added that polls show young people now view unauthorised AI training as theft. The RIAA's new strategy is therefore to first defend the basic rights of creators, and then let fans make their own choices in the open market. As Mitch put it, fans will always win in the end, because every service eventually relies on people paying for subscriptions. But this logic has a prerequisite: fans have to know exactly what they are listening to.

Deepfakes take this problem a step further. Mitch pointed out that there is no existing U.S. federal law granting ordinary people a universal right to their own voice and likeness, and the right of publicity under state law usually only covers commercial ads and endorsements. You currently have no legal standing to stop someone from using your face to make non-commercial content that has no connection to product endorsement. The NO FAKES Act is designed to fill this legal gap, creating a new form of intellectual property that works similar to trademark, while learning lessons from the DMCA: it sets thresholds for issuing takedown notices, penalizes false claims, creates clear mechanisms for counter-notices, leaves First Amendment issues such as parody and news reporting for courts to review quickly, and forbids platforms from ruling on constitutional questions themselves. The key detail here is that this creates a new layer of rights entirely separate from standard copyright law.

The most interesting segment of the interview touched on the internal divide at Google. YouTube frames its content ecosystem as a licensed system, and music is the lifeblood of YouTube. That is why even after YouTube won the high-profile Viacom lawsuit, it still built the Content ID system, forming a symbiotic relationship with the music industry. Google Search and the Gemini AI team, by contrast, are far more willing to frame model training as fair use, scraping large volumes of data first and fighting out the legal battles later. Mitch summed it up by noting that YouTube realized you could win a court case and still lose your entire business. Music is directly consumed by fans, and the experience of human music is irreplaceable; search results only point users to existing works, so the commercial logic is completely different. This observation applies equally to all AI companies: training data will eventually turn into concrete commercial constraints, which determine whether a product can consistently supply high-quality content in the long run.

Labeling is another line of effort the RIAA is pushing right now. Mitch said Spotify's current practice of only marking AI content in the credits section is not enough: labels should be required to mark AI content at the individual track level. Detection technology is already mature for fully AI-generated music, and Deezer's current false positive rate is very low. The real bottleneck sits in the distribution chain: bedroom producers will not voluntarily fill in the required metadata, so distributors such as DistroKid and TuneCore act as the actual gatekeepers. California's transparency law requires Suno to add machine-readable watermarks to its outputs starting in August, and this regulation will likely push global platforms to adopt unified standards, since implementing geo-blocking for a single country is far more expensive than rolling out compliance across all global markets. This is a very realistic calculation: regulatory pressure and market pressure will work together to drive change.

The RIAA only has 46 employees. Last year, the U.S. recorded music industry generated $11 billion in wholesale revenue, while a single AI company can hold a market cap of $2 trillion. Mitch makes his decisions by actively seeking out opposing views, asking friends working on the other side of the debate: "How would you argue against me?" He sticks to roughly 60% of his initial instincts, while adjusting the other 40% of his plans based on new feedback. The music industry's certification system is also changing: 1500 streams now equal one certified unit, and fully machine-generated content will almost certainly be excluded, but the line for what counts as "substantially human" will keep shifting as AI-assisted creation becomes more widespread.

Mitch closed by noting that the music industry went from wax cylinders to AI in only a hundred years, and it is far more adaptable than most other industries. Nilay added that the music industry is a preview of the future for all creative sectors. This is no exaggeration: AI copyright conflicts will not stay limited to music, it is just that music is the first creative field to be fully scanned for training data. The real variable sits in the middle layer: platforms, distributors, labels, licensing agreements, and whether fans are willing to pay a premium for "human-created" work. Fans will make their own choices, but they need accurate information to make those choices. Clear labeling and clear rights are the foundation for that information.