Volume 2, Issue 1
will.i.am gives a frustrating interview
We kicked off the year with will.i.am declaring that AI heralds a new “Renaissance” for creatives. In a Bloomberg interview, he says he is more worried about his accountant than his drummer. He seems to want to avoid saying that AI would become prevalent in music, although he did kind of let go at the end on it. He argues that just as we distinguish between “organic” oranges and regular ones, we will soon have to distinguish between organic human music (which he calls “lived experience”) and AI music. Honestly, I was a little disappointed, as I think he missed a golden opportunity to make stronger statements about artists’ rights, labeling, and ethical AI.
The message about real experience is one I do agree with, although I align more with Veronica Hylak’s recent take. She argues that real music is about breaking the rules—pushing the tempo in a big section or changing volume instinctively. AI tends to play perfectly on the grid; it doesn’t know how to interpret silence or “struggle.” She joins a chorus of folks (Scott Belsky comes to mind) who talk about how the value of a human performance will grow as AI dominates.
Legal Briefs: Elon Musk vs. The Publishers
Just when you thought the lawsuits were settling down, Elon Musk’s X Corp has sued the major music publishers and the NMPA alleging that the industry “colluded” to force X into paying “supra-competitive” licensing fees by inundating the platform with DMCA takedowns.
Musk claims the publishers coordinated a “boycott” to force a license. To most of us in the industry, refusing to use a service that doesn’t pay for your IP isn’t collusion; it’s a perfectly reasonable response. As “Top Music Attorney” Miss Krystle points out, X is arguably the only major social platform that hasn’t secured these licenses. So they are effectively suing the people whose work drives their platform, and it illustrates how 2026 will be another year of continued high-stakes litigation.
C2PA & Authenticity:
I’m very pleased to announce that the C2PA released version 2.3 of the C2PA specification. I am most active in the audio and watermarking task forces of the C2PA and super excited to have been a part of this release, bringing more audio goodness to the specification. 2.3 contains modest updates for supporting different file types, including expanded support for OGG Vorbis as well as further refinements for embedding manifests in other audio formats.
2.3 also introduces new actions specific to audio processing, as opposed to video and images, which are more fully represented in the specification. These additions are c2pa.mixed, c2pa.mastered, and c2pa.remixed. Many thanks to the task force leaders as well as everyone at the C2PA for delivering this important work. These new labels ensure that audio creators using C2PA can better describe the specific production state of an asset using industry-aligned language.
There’s a lot more goodness coming in 2026, so stay tuned…
And now, let’s talk about the elephant in the room with regard to audio watermarks. A new paper has revealed significant weaknesses in current audio watermarking generated by AI music models. Researchers found that while watermarks work in isolation, they are failing against realistic attacks like the kinds of processing used in mastering chains, by DJs, and in aggressive bus compression. The key finding is pretty, um, conclusive:
Key Finding. This work reveals that none of the existing audio watermarking techniques are fully reliable against all forms of attack.
This mirrors my own experience. I have done a fair amount of testing on watermarks, including doing lots of aggressive mixing, as one might find out in the wild. Chopping out sections, changing the pitch, the speed, adding effects, stemming, and recombining. Audio watermark performance is different across a lot of these tests, but I maintain that it is relatively easy to damage a watermark through very simple audio mixing processes. I would bet that if you took any watermarked file, sped it up by 30%, then slowed it down by 30%, you’d break most watermarks and not be able to audibly hear a difference without a controlled experiment.
Now, a reality check. This is not a reason to not use watermarks! Watermarks are only one of the pillars of provenance, you know? In the C2PA world, there is the sealed manifest in the metadata, an inaudible watermark with soft binding to said manifest, and even the potential for a fingerprint. So, lots of layers.
Platform Updates: Licensed AI is Here
While some platforms are fighting the labels, Hangout is taking an artist-centric approach. They launched a generative AI feature called Turntable.AI that is fully licensed across labels and publishers. CEO Joseph Perla calls it the “ChatGPT moment for music, but made legal.” It’s a stark contrast to the scrapers: every generation pays rights-holders, and every training set is cleared.
The Final FreakOut
Closing out the month, we get the news that Native Instruments and associated brands, including iZotope, have gone into bankruptcy proceedings. This is causing quite a bit of consternation because so many of us rely heavily on these tools, and particularly in mixing and mastering, they are AI-driven. And really, really fantastically good. So we all have our fingers crossed. There are lots of takes on this, but I’ll link to Dave Kropf over at 52 Cues, who covered the initial announcement and then a follow-up after further statements by Native Instruments.
Closing Thoughts
2026 is going to be a big year for the music industry and Catamount Music. Stay tuned for some incredibly exciting projects!
Leave a Reply