
.png)
The European Parliament voted 460-71 this week to force generative AI companies to respect copyright law, disclose what they trained on, and pay creators for using their work. It's a big deal. Not because it's law yet (it isn't), but because a 460-71 margin tells you which way this is heading.
For anyone who's been watching the AI and content space, this isn't surprising. The creative industries have been getting quietly hollowed out for two years while AI companies trained on everything they could find. The Parliament is essentially saying: that era is ending.
AI providers must disclose every copyrighted work used in training, bear legal costs when infringement cases go against them, and give rights holders real opt-out mechanisms. News organizations get compensation when AI systems divert their traffic and revenues. AI-generated content cannot receive copyright protection. Global flat-rate licensing deals are explicitly rejected.
What's interesting is that the framework they're describing isn't just about paying creators after the fact. It requires technical opt-out mechanisms: a way for rights holders to signal, at the content level, that their work isn't available for training. That's not a legal document or a terms of service update. That's infrastructure.
This is exactly what Limbo does. Every piece of content that goes through Limbo carries a signed provenance record with a Do Not Train assertion embedded in the file itself. It travels with the content wherever it goes. It doesn't disappear when a platform strips metadata, because it's not just metadata. And when an AI company ingests content that carries that signal, there's no longer a credible "we didn't know" defense.
The Parliament rejected flat-rate global licensing deals, which means the remuneration framework they're imagining is more granular: individual rights holders, individual terms. That requires machine-readable licensing signals embedded at the asset level. C2PA manifests can carry exactly that. It's something we're actively thinking about at Limbo.
The vote isn't binding legislation. But the direction is clear, and organizations that build content rights infrastructure before the deadlines arrive will be in a very different position than those scrambling to retrofit it afterward.
If you want to talk through what that looks like for your organization, we're here.
.png)