220+ Rights Groups Tell France: Make AI Companies Prove They Didn't Steal Your Music

The Darcos bill would flip the burden of proof in AI copyright cases — and the global creative industry is lining up behind it.

Written by Grant Hollis · Bureau Media · Jun 9, 2026

The Problem With the Current Setup

Right now, if you're a songwriter, a label, or a publisher who suspects an AI company trained its model on your catalog, the legal burden is yours. You have to prove it happened. Given that AI training datasets are typically undisclosed, proprietary, and vast — we're talking billions of files — that's roughly equivalent to being asked to prove someone read your book without ever seeing their library.

That's the structural absurdity the Darcos bill is designed to fix.

What the Bill Actually Does

The legislation, named after French senator Loïc Hervé Darcos, would invert the evidentiary standard in AI copyright disputes under French law. Instead of creators bearing the burden of proof, AI companies would be required to demonstrate that copyrighted content was *not* used in their training data.

This is a meaningful legal distinction. It transforms transparency from a voluntary PR gesture into a legal obligation with real consequences for non-compliance.

220+ Organizations and Counting

The coalition backing the bill spans more than 220 rights organizations worldwide — a number that reflects how broadly the creative sector has aligned on this issue. Music publishers, collecting societies, performer rights groups, and labels across multiple continents have added their names, turning what could have been a French domestic debate into a coordinated international lobbying effort aimed at a single legislative chamber.

The message is deliberate: France has an opportunity to move first, and the rest of the world is watching.

Why France, Why Now

France has historically been one of the more assertive European jurisdictions on intellectual property, and its legislative calendar has created an opening. The Darcos bill is positioned as a targeted intervention — not a sweeping AI regulation, but a specific procedural correction to how copyright disputes are adjudicated when AI training is the alleged infringement.

That narrowness is probably strategic. Broad AI regulation tends to get bogged down in competing interests and implementation timelines. A burden-of-proof flip is a surgical move that works within existing IP law rather than trying to rewrite it.

The Bigger Stakes

If France passes this, the compliance math for AI developers changes immediately — at least for any company operating in or selling into the French market. More importantly, it creates a legal precedent that other jurisdictions can reference and replicate.

The creative industry has spent the better part of three years watching AI companies build products on content they didn't license, then argue in court that proving infringement is the other side's problem. The Darcos bill is the clearest legislative attempt yet to close that gap. Whether France's National Assembly moves on it will say a lot about how seriously governments are willing to take the structural imbalance between AI developers and the people whose work trained them.

Key takeaways

The Darcos bill would reverse the burden of proof in AI copyright disputes in France, requiring AI firms to show they didn't use protected content — not the other way around.
More than 220 rights organizations from around the world have signed on in support, signaling that this is a coordinated global push, not a local French policy debate.
Current copyright frameworks put the onus on creators to prove their work was ingested — a near-impossible task given the opacity of AI training datasets.
France's National Assembly is the legislative target, making this a test case for how democratic governments can regulate AI training practices through existing intellectual property law.
A successful Darcos bill could set a template that other jurisdictions adopt, raising the compliance cost for AI developers who have so far operated with minimal transparency obligations.

FAQ

What is the Darcos bill?

The Darcos bill is French legislation that would reverse the burden of proof in AI copyright disputes. Under the bill, AI companies would be required to prove that copyrighted content was not used to train their models, rather than creators having to prove it was.

Why does reversing the burden of proof matter?

AI training datasets are typically opaque and undisclosed, making it practically impossible for individual creators or rights holders to prove their work was included. Shifting the burden to AI companies forces them to maintain and produce records demonstrating compliance — a transparency obligation they currently don't face.

Who is supporting the bill?

More than 220 rights organizations worldwide have urged France's National Assembly to adopt the legislation, including music publishers, collecting societies, and performer rights groups across multiple countries.

Could this affect AI companies outside France?

Any AI company operating in or selling services into the French market would need to comply if the bill passes. Beyond direct legal exposure, a French precedent could influence similar legislation in other jurisdictions, raising the global compliance stakes for AI developers.

Has any country already passed similar legislation?

As of the time of this report, no jurisdiction has enacted a law specifically reversing the burden of proof in AI copyright disputes in this way. The Darcos bill, if passed, would be a first.