YouTube wants to make AI Drake fakes official

Earlier this year, an AI-generated fake track featuring Drake and The Weeknd went viral and terrified the music industry into firing off takedown notices. Now, YouTube wants to make that exact voice-cloning experience an official, monetized product.

According to reports from Bloomberg and Billboard, YouTube is currently building a generative AI tool that lets creators make videos and songs using the cloned voices of famous recording artists. The company is actively in talks with major labels to license music catalogs for model training.

YouTube's play for legit voice clones

YouTube had actually hoped to unveil the tool back at its Made On YouTube event in September, alongside its AI-generated background and video ideation features. The voice cloner missed the stage simply because licensing negotiations dragged on.

Under the current beta blueprint, a small group of artists who opt in will license their vocal likeness to a handpicked pool of creators. Only if that trial succeeds would YouTube consider rolling it out broadly across the platform.

Beyond just testing the tool, YouTube is also asking these participating musicians for input to help steer the platform's broader generative AI strategy going forward.

Labels want innovation, but artists have cold feet

All three major labels, Universal Music Group, Sony Music Entertainment, and Warner Music Group, are at the table, but the talks are proving tricky.

Label heads are terrified of repeating the mistake they made two decades ago, when resistance to digital downloads and streaming cost the industry dearly. Warner CEO Robert Kyncl and UMG chief Lucian Grainge have both spoken publicly about embracing AI.

However, some executives privately worry they have surrendered leverage by acting too eager to appear innovative. Persuading top-tier artists to join has also been an uphill battle. Outside of outlier figures like Grimes, most major artists, including Sting, John Legend, and Selena Gomez, remain deeply uneasy about unvetted creators putting words in their mouths.

Then comes the legal headache of payouts. Should artists get paid based on the training data fed into the model, or based on the output and views the AI generated? Publishing rights are even messier, with hundreds of songwriters tied to any given catalog. Billboard reports YouTube may prefer paying a single lump-sum fee to bypass the publishing nightmare.

Why it matters

For everyday creators, having legitimate access to iconic voices could unleash an entirely new wave of audio-visual remix culture. It turns anyone into an A-list vocalist, though the initial beta will likely remain an exclusive playground for top creators.

We think YouTube is playing this very smartly. Instead of scraping songs and waiting for lawsuits like many generative AI startups do, YouTube is leveraging its Content ID reputation to position itself as the responsible industry gatekeeper. If YouTube can establish the blueprint for how AI voice licensing and revenue splits work, competitors will have to follow its lead.

That said, the reputational risk for artists remains the biggest roadblock. Music labels might want to monetize AI, but stars protect their likeness above all else. Unless YouTube builds airtight guardrails that give artists total control over context and lyrical content, convincing top-charting talent to willingly hand over their vocal identity will be a brutally tough sell.