AI has moved from the margins of music production to the center of it, at least according to Harvey Mason Jr., the CEO of the Recording Academy, the organization behind the Grammys. In a wide-ranging conversation about technology, creativity, and the future of awards eligibility, Mason described generative AI as “omnipresent” in studios—so common, he said, that he can’t remember a recent session that didn’t involve some form of AI.
That shift is forcing the Grammys to confront a problem that’s both practical and philosophical: how do you recognize excellence when the tools used to create music are changing faster than policy can keep up? And how do you do it without turning the awards into a popularity contest for synthetic output—or, worse, diluting the meaning of “human creativity” that the Grammys have long tried to honor?
Mason’s answers, delivered with the perspective of a working producer and songwriter as well as an executive responsible for the Recording Academy’s rules, suggest the Grammys are not planning a simple ban on AI. Instead, they’re trying to draw a moving line—one that depends on category, on the amount of human involvement, and on whether the system can reliably verify what happened in the creative process.
The result is a policy framework that Mason describes as a “tightrope,” because the Academy can’t currently measure AI contribution with the precision people might want. It has to rely on disclosure, screening committees, and documentation—plus the assumption that the music community will behave like an “honorable community,” even as the volume of AI-generated material grows.
What AI is doing in real studios, not just in demos
Mason’s most concrete description of AI’s impact came when he talked about what he sees in sessions, particularly in pop and R&B, where he says AI use is especially widespread. He doesn’t frame AI as a single replacement technology. Instead, he describes it as a set of capabilities that show up in different parts of songwriting and production.
In his account, AI is being used to generate chord progressions, fill out drum loops, and—sometimes—create entire tracks. It’s also being used to support lyrics: not necessarily to write everything from scratch, but to help with second verses that match the rhyme scheme and rhythm of the first, or to generate ideas based on a title. Mason even characterized some usage as a kind of “rhyming dictionary,” where creators prompt for alternatives rather than starting over.
Then there are the production workflows that are less visible to listeners but central to how songs get built. Mason said AI is used to create background vocal stacks, generate demos of singers for artists who may be writing for them, and assist with other studio tasks that traditionally required time, iteration, and specialized skill.
The key point is that AI isn’t only producing finished songs. It’s increasingly acting like a collaborator inside the process—offering suggestions, generating rough drafts, and accelerating the early stages of composition. That’s why Mason’s reaction is mixed. He’s impressed by the quality improvements he’s seen over the last 18 months, but he’s also disturbed by what those improvements imply for the value of effort and craft.
He described a moment that captures the speed of change: 18 months ago, he said it was easier to tell when something was AI-generated. Now, he’s hearing people play him tracks and telling him AI made them—and he’s surprised. The implication is that AI output is becoming harder to distinguish, which complicates both consumer perception and award eligibility enforcement.
The Grammys’ core rule: AI isn’t automatically disqualifying, but human creativity must lead
The Recording Academy’s rules, as Mason explained them, are designed around a principle: AI use doesn’t automatically make an artist ineligible. But the Academy wants to ensure that human creativity is at the forefront of what’s being recognized.
This is where the “de minimis” language comes in. Mason said the current threshold is “more than a de minimis amount of human creativity involved in the process.” In plain terms, the Grammys are looking for meaningful human contribution—not just a tiny interaction where a creator prompts a system and then takes the output as-is.
But he also emphasized that the system isn’t perfect. The Academy doesn’t have a black-and-white box it can check that precisely proves how much human work occurred. That’s why the policy is described as a tightrope: it has to be strict enough to protect the meaning of the awards, but flexible enough to reflect how AI is actually being used in studios today.
Mason offered examples of how category boundaries might work. If AI is used for background vocals or other elements, that doesn’t necessarily eliminate eligibility across the board. He said someone could still submit for songwriting categories if the human role is substantial enough. Conversely, if AI is doing the “performing,” then performance awards become the problem—because the Grammys are trying to honor human performance.
He also described a scenario that illustrates the logic: if AI wrote the song but a human sang it and “sang the heck out of it,” then the human performer could potentially be submitted for a performance award. The focus is not on whether AI existed somewhere in the production chain; it’s on whether the human contribution is meaningful in the specific aspect being awarded.
This approach reflects a broader reality: AI is being used in many ways, and the Grammys are trying to avoid a blanket rule that would punish legitimate hybrid workflows. At the same time, Mason acknowledged that some uses seem “egregious and too much,” implying that the Academy will continue to refine what counts as acceptable.
How the Grammys try to prove what happened
One of the hardest questions in AI eligibility is verification. Mason said the Academy asks submissions whether AI was used, but he also noted that the system is partly dependent on creators disclosing accurately. He compared the situation to “Ozempic,” referencing the idea that everyone knows it’s being used but people don’t always talk about it openly.
The Academy’s screening committees review claims and evaluate evidence. Mason said this can involve asking questions, requesting documentation, and relying on analysis. But he was clear that the process isn’t a perfect science. There isn’t currently a reliable, universally deployed technology that can determine when AI was used and how much of a track was generated by AI.
So the Grammys are left with a combination of disclosure, committee review, and—until better detection exists—trust. Mason argued that creators generally don’t want to cheat and win on grounds they can’t prove, and he suggested the music community is likely to be forthcoming. Still, he admitted the system is challenging because it’s difficult to determine exactly who did what.
That difficulty becomes more acute as AI output becomes more convincing. If AI can produce high-quality material that looks indistinguishable from human work, then the burden of proof shifts toward documentation and disclosure rather than technical detection.
Will the Grammys get overwhelmed by AI volume?
Another practical concern is scale. Mason said the Academy had about 24,000 submissions last year, and that number has increased compared to earlier years. He didn’t claim the Grammys are currently inundated by AI entries, but he warned that if AI-related submissions grow enough to dilute evaluation or overwhelm the process, the Academy would have to make changes.
This is an important nuance: the issue isn’t only whether AI is present. It’s whether AI increases the volume of eligible material to the point where the system can’t evaluate it meaningfully. Mason framed this as something the Academy is actively watching, listening to, and preparing to adjust.
In other words, the Grammys aren’t just reacting to AI as a cultural controversy. They’re treating it as an operational variable that could force procedural evolution.
The public gap: people use AI, but fans may not want to hear it
Mason also addressed the mismatch between creator behavior and audience sentiment. He referenced polling that suggests many people don’t want to listen to music made with AI help, and he noted that younger audiences may be more resistant even if they use AI more themselves.
But he also pointed out that the Academy’s job is complicated by the fact that creators may not disclose AI use in a way that aligns with audience preferences. He said the Academy is “at the mercy” of people telling them and disclosing when they’re using it, until detection technology improves.
He didn’t predict whether the public gap will close or widen. Instead, he offered a historical pattern: societies often become comfortable with new technologies once they’re normalized. He compared it to earlier reactions to tools like Pro Tools, AutoTune, and Melodyne—technologies that were once controversial but became standard.
Still, Mason’s optimism is tempered by the fact that AI is not just a production tool. It’s also tied to ownership, authorship, and the authenticity of creative labor. That’s why the debate feels different from earlier waves of music technology.
The Grammys’ dilemma is not only artistic—it’s economic and legal
While Mason focused heavily on eligibility rules, his comments also made clear that AI is entangled with broader issues of compensation and intellectual property. He argued that creators should be compensated and should have control over how their work is used, credited, and monetized.
He discussed legislation aimed at protecting voice, image, and likeness (the No Fakes Act), improving access to training data records (the TRAIN Act), and increasing transparency (the CLEAR Act). He said there is bipartisan alignment around the No Fakes Act, but he didn’t expect Congress to act quickly.
He also described the current environment as one where platforms invent private frameworks. He referenced YouTube’s likeness detection and Content ID-like approaches, suggesting these steps can be a “first step” but also create friction for artists because they lack a unified federal framework.
This matters for the Grammys because eligibility rules are only one part of the ecosystem. If creators can’t control how their voices and likenesses are used, then the question of what counts as “human creativity” becomes harder to answer ethically, not just administratively.
Mason’s