Взлом Suno раскрыл: AI-генератор музыки скрапил YouTube, Deezer и Genius

A hack has revealed that AI music generator Suno was trained on decades of music and podcasts scraped from the open internet — including YouTube, the streaming service Deezer, and the lyrics site Genius. This is reported by 404 Media, which examined the service's leaked source code.

What the hack revealed

Suno's leaked source code directly points to the sources of its training data: it contains references to YouTube, Deezer, and Genius. This is the first technical confirmation of exactly where the service got the music and podcasts to train its models — previously the company kept its list of sources secret.

Suno is an AI music generator that creates finished tracks with vocals and instrumentals from a text description. According to 404 Media, the scraping covered "decades" of audio recordings from the open internet.

• The data source was YouTube (video and audio), Deezer (streaming), and Genius (song lyrics)
• The data included not just music, but podcasts as well
• The disclosure came through a source code leak, not an official statement from the company
• The outlet that revealed the finding is 404 Media

"The hacked source code shows how

Suno scraped decades of music and podcasts from the internet to train its AI tool," the 404 Media investigation states.

Why this matters for the lawsuits

The finding hits Suno at a moment of litigation with the music industry. In June 2024, the Recording Industry Association of America (RIAA), on behalf of Universal Music, Sony Music, and Warner Records, filed a lawsuit against Suno, accusing the startup of mass copyright infringement during training.

Suno had previously effectively acknowledged the scale of its data collection. In its legal response, the company stated that its training dataset "includes essentially all music files of reasonable quality available on the open internet," and insisted that such training falls under the fair use doctrine. The leaked code gives plaintiffs concrete names of platforms, strengthening their position.

What it means

The debate over what it is legal to train generative models on is moving from the realm of statements into the realm of technical evidence. The leak of Suno's source code shows that the question of training data sources for AI companies is ceasing to be a black box — and this concerns not just music, but the entire generative AI industry.

Frequently asked questions

What is Suno?

Suno is an AI-powered music generation service: from a text prompt, it creates finished songs with vocals and instrumentals. The data used to train its models became the center of a scandal after the source code leak.

What is the lawsuit against Suno about?

In June 2024, the RIAA, on behalf of Universal, Sony, and Warner, accused Suno of training its models on copyrighted music without rightsholders' permission. Suno insists that training on open data is fair use.