How AI-Generated Music Got Here: The Milestones That Changed Everything

If you want to understand the fight over AI music happening right now, you have to rewind to a weekend in April 2023 — the moment the whole thing stopped being a research demo and became everyone's problem.

This is a plain-English timeline of how we got here: the songs, the lawsuits, the first arrest, and the legal scaffolding that's still being built around AI-generated music. No hype, no doom — just the milestones that actually mattered.

The short version

• April 2023 — "Heart on My Sleeve," a fake Drake/The Weeknd song made with AI, goes viral and is pulled within days. The industry panics.
• 2023–2024 — Suno and Udio turn AI song-making into a consumer product anyone can use.
• June 2024 — The major labels sue both companies for copyright infringement.
• July 2024 — Tennessee's ELVIS Act becomes the first US law to protect a person's voice from AI cloning.
• September 2024 — The first criminal case: a musician charged with using AI songs and bots to steal $10M in streaming royalties.
• January–May 2025 — The US Copyright Office and Congress start drawing the legal lines.

Now the longer story.

April 2023: the fake Drake song that started it all

A track called "Heart on My Sleeve" appeared online, credited to a then-unknown creator named Ghostwriter. It used AI-generated voice models to mimic Drake and The Weeknd so convincingly that millions of people shared it without realising — or caring — that neither artist had ever set foot in the studio. Across TikTok, YouTube and streaming platforms it pulled tens of millions of plays in a matter of days.

Then Universal Music Group leaned on the platforms, and the song vanished almost as fast as it arrived.

That single weekend reframed the entire conversation. It wasn't a future hypothetical anymore. You could clone a global superstar's voice, write a passable hit, and reach a stadium-sized audience overnight — and the people who built their careers on those voices had almost no idea how to stop you. The saga got stranger before it ended: later in 2023, "Heart on My Sleeve" was even briefly submitted for Grammy consideration before the Recording Academy ruled it ineligible.

2023–2024: Suno and Udio make it everyone's tool

The deepfake scared the industry. What changed it permanently was accessibility.

Two companies — Suno and Udio — turned AI music generation into something your non-technical cousin could use. Type a prompt like "melancholy indie folk about a long drive home," wait twenty seconds, and you'd get a finished song: vocals, instruments, structure, hook. No instrument, no studio, no skill required.

The viral moments piled up. In 2024, a comedic track called "BBL Drizzy," made with Udio, blew up to the point that Metro Boomin released a free beat over it. Suddenly AI music wasn't just imitation — it was generating original-ish songs that real people genuinely wanted to listen to and remix. The technology had crossed from novelty into something that looked, uncomfortably, like a creative tool.

June 2024: the labels go to war

The major labels stopped issuing statements and started filing lawsuits.

In June 2024, the RIAA — acting for Universal, Sony and Warner — sued both Suno and Udio for copyright infringement, accusing them of training their models on enormous libraries of copyrighted recordings without permission or payment. The labels wanted damages that could run into the billions.

The AI companies' defence boiled down to one phrase you're going to keep hearing: fair use. Their argument: learning from existing music to generate something new is transformative, the same way a human musician learns by listening. The labels' counter: this isn't a person being inspired — it's a commercial product copying their catalogue wholesale to compete directly with the artists it copied.

That tension — inspiration versus ingestion — is the heart of the entire debate, and it still hasn't been resolved.

July 2024: a voice becomes property

While the copyright fight played out, lawmakers went after a different angle: your voice.

On July 1, 2024, Tennessee's ELVIS Act (Ensuring Likeness, Voice and Image Security Act) took effect — the first US law to explicitly treat a person's voice as a protected property right against AI cloning. Named, of course, with Memphis in mind, it gave artists a concrete legal tool against the kind of impersonation that "Heart on My Sleeve" pioneered.

It was a state law, not a national one. But it set a template, and it signalled that "you can't just clone someone's voice" was about to become more than an ethical opinion.

September 2024: the first criminal case

Then came the milestone that revealed the grubby commercial reality underneath the artistic debate.

In September 2024, US prosecutors charged a North Carolina musician, Michael Smith, in what they called the first criminal case of AI-assisted streaming fraud. The scheme was audacious: generate hundreds of thousands of AI songs, then use armies of bots to stream them billions of times, quietly collecting the royalties. Spread thinly across enough tracks, the fake streams looked normal. Prosecutors say it netted him more than $10 million.

This was the part nobody wanted to talk about. The threat from AI music wasn't only that a robot might write a better chorus than you. It was that the streaming economy — already paying fractions of a cent per play — could be flooded with infinite cheap content designed purely to siphon money out of the same royalty pool real artists depend on.

January–May 2025: the rules start to take shape

By 2025, the institutions finally started drawing lines.

In January 2025, the US Copyright Office published a major report on copyrightability, landing on a clear principle: purely AI-generated output isn't protected by copyright, because copyright requires human authorship — and typing a prompt doesn't make you the author. Music that blends genuine human creativity with AI tools can still qualify. Pure push-button output can't.

In April 2025, the bipartisan NO FAKES Act was introduced in the Senate — an attempt to do nationally what Tennessee did locally: make unauthorised AI replicas of someone's voice or likeness illegal across the country, with carve-outs for satire, news and commentary.

And in May 2025, the Copyright Office released its long-awaited report on AI training — the core of the Suno and Udio lawsuits. It pointedly rejected the idea that training on copyrighted work is "inherently transformative," and warned that it threatens real harm to the market for human-made music. (In a twist that tells you how high the stakes have become, the head of the Copyright Office was dismissed the very next day.)

Where this leaves us

In barely two years, AI music went from a viral curiosity to a phenomenon with active billion-dollar lawsuits, a criminal conviction, new state laws, federal bills, and the Copyright Office staking out its position. The novelty phase is over.

What hasn't been settled is the hard question underneath all of it: when a machine learns from a lifetime of human music and produces something people want to hear, who deserves the credit — and the cheque?

We have opinions about that, because Videojam is built on the opposite bet: that the music worth your time is the music a human chose to point at. The next chapter — the lawsuits getting messy, the deepfakes getting better, and the streaming platforms scrambling to respond — is where the real debacle begins.

This is the first in our ongoing AI & Music series, tracking how generative AI is reshaping the music people actually listen to.