How is the industry adapting to the AI era?

When I launched this newsletter, I struggled to find topics worth writing about. Back then, serious projects with the potential to set precedent and add long-term value were scarce. Since July 31, that's changed. My job now is picking which of the dozens of developments unleashed by the first-instance ruling in GEMA v. Suno to cover.

Today I want to focus on two EU rules that are now enforceable and that, in my view, will have a positive impact on the creative ecosystem. I also picked out a significant policy shift from KOMCA, South Korea's music rights society, in how it handles AI-generated works.

Both feel like concrete evidence of an industry learning to adapt to the AI era.

The European Union's rules

The AI Act crossed two milestones at the same time on August 2. The labeling rules for AI-generated content took effect, and the AI Office gained the power to enforce a second block of training-related obligations that had already been in force since August 2025.

The first is about labeling AI output. Providers of these systems now have to mark their outputs in a machine-readable way, using watermarks or embedded metadata. In other words, a technical signal that lets another system detect the content's origin.

What's interesting is that this label lets major platforms (Instagram, X, YouTube, TikTok) satisfy an obligation they already have under the DSA, the EU's other major digital law. The DSA requires them to identify deepfakes and flag them visibly to users. Without a signal like the one the AI Act mandates, catching every deepfake by hand is close to impossible. With it, it becomes far more workable.

Here's a real-world example to make this concrete. Say you're scrolling social media and come across a controversial video of your country's president. The AI system that generated it has already tagged it with that invisible technical signal. The platform detects it, and now it can, and must, show you a clear notice that the video was AI-generated.

Personally, I think this rule is entirely necessary. It's getting harder, even for a trained eye, to tell what's real from what's fake. Misused, this technology has the potential to manipulate public opinion at an unimaginable scale.

The other block of rules governs how models train on copyrighted content. It had already been in force since August 2025. What changed is that the EU can now actually enforce it.

Model developers have to publish a sufficiently detailed summary of the content used for training, in a format defined by the AI Office. The obligation applies to any model offered in the EU, regardless of where it was trained, so training outside the bloc to dodge the rule isn't a way out.

Alongside that, the same developers need a policy in place to respect European copyright and the rights reservations rightsholders file under the opt-out mechanism, already used by Sony Music and Warner Music Group.

Fines for breaching either block reach €15 million or 3% of global annual revenue, whichever is higher.

The KOMCA case

Looking into KOMCA, I was struck by how early it engaged with this issue. On March 24, 2025, it rolled out a zero-tolerance rule for registering works that used AI anywhere in their process. At the time, KOMCA was already watching AI-made music grow fast, with no legal standard yet in place for how to handle it. The requirement was 100% human contribution.

One thing worth clarifying. KOMCA administers copyright over compositions and lyrics, not rights over recordings. That distinction matters, because some Korean companies have been using AI in recording work without touching composition rights. HYBE, for example, released a MIDNATT track simultaneously in six languages using voice technology, without any of that going through KOMCA.

The society is now lifting the ban, but only for cases where a human played a substantial, leading role in the melody, lyrics, structure, or arrangement. Despite that loosening, registration now demands considerably more than before. Anyone registering an AI-assisted work has to declare which parts involved AI, which tools were used, and what their own contribution actually consisted of. KOMCA can request evidence of the process, run technical checks, and put the case through internal review.

The consequences for lying are harsh too. The society can withhold royalties, claw back what's already been paid, and terminate the creator's management contract. If there's intent or gross negligence, it can also claim back up to three times the amount wrongly collected.

This shift is in line with the U.S. Copyright Office's stance. Writing prompts doesn't make anyone the author of the output, and purely AI-generated material has to be declared and stays outside any copyright claim. AI can help make a song, but it can't sign it.

While I think the GEMA v. Suno case played a role in this policy shift, there's more to it. Back in March of this year, South Korea's auditing body published a review of AI readiness covering eleven of the country's collective management organizations. As part of that work, it sampled 8,540 songs registered by 29 high-volume songwriters and estimated that 60.9% had likely used AI. The number made clear that these societies had been registering works and paying royalties without checking whether AI was involved at all.

Closing thoughts

When I first started studying these issues, I'll admit it felt premature to talk about regulation. We simply didn't have enough experience to get it right. We're now entering a different phase, one where standards are starting to take shape that will benefit everyone, including the tech sector itself.

As a metaphor, I keep coming back to blockchain and crypto, a technology that was born with a rebellious, anti-regulation spirit. Over time, it became clear that it needed clear rules to actually function. Today the crypto ecosystem is much quieter and has found points of contact with traditional finance that would have seemed unthinkable ten years ago. I think the same thing is going to happen with generative AI. Clear regulation, arriving at the right time, will let the creative sector keep growing alongside it and put to rest the fear of total replacement.