SZA calls out ‘Suno’ and musicians using AI

Suno’s Approach to Platform Integrity and Music Copyright Safeguards

Last week, we shared that we’re now testing a new music model developed in partnership with the music industry.

We’re extremely excited about this, and it’s a good opportunity to share more on how we think about platform integrity and protecting the people behind the music.

At Suno, preserving artistry and human creativity is just as important as product innovation itself. We know that’s easy to be skeptical about – especially if you haven’t actually immersed yourself in our product. But most of the people building Suno are musicians. Empowering human creativity is why we show up, not just some corporate positioning. We believe companies building AI have a responsibility to invest in safeguards with the same level of ambition and rigor that they bring to building products.

From our earliest days, we worked to build protections directly into the foundation of our platform.

That includes clear rules prohibiting users from uploading or distributing content they don’t own or have the rights to use, meaningful enforcement when those rules are violated, and partnerships with industry-standard providers like Audible Magic, Musixmatch, and ACRCloud to help us identify and prevent misuse.

It’s also shaped how we’ve approached model development itself.

One of the most common questions people ask about AI is whether models can reproduce material from their training data. Our answer is simple: no, that should not happen. Our philosophy has always been that AI should help people create new music, not replicate someone else’s. That’s why we built our models around what we call “Original Creation, By Design,” training strategies intended to reduce the risk of generating unauthorized reproductions.

For example, we do not use artist names as a category of training metadata – meaning we made an intentional choice not to teach models artist names because our goal is to help people create brand new songs, not music that sounds like existing artists.

Alongside those efforts, we’re actively developing additional approaches for audio fingerprinting, watermarking, and spam and impersonation detection. In addition, we’re engaging in discussions around AI labeling with our users and creatives at all levels, to understand how to best empower them while respecting privacy.
 
We don’t treat these issues as a compliance exercise. It’s a core part of how we build and it’s led us to put more copyright safeguards in place than traditional music software. Not just compared to AI music tools, but compared to all the music tools that preceded them.

We’re in active dialogue with artists, songwriters, producers, labels, distributors and partners across the music industry, and those conversations will continue to shape the safeguards we build and the investments we make going forward. This will become increasingly important as we release our partnered models in the coming months.

We’ll have a lot more to share soon. | 31 comments on LinkedIn