AI copyright talks shift from revenue charges to usage-based royalties

In short: There is no clear evidence of a single “revenue-based charge” policy for AI today, and 2026 discussions are instead focusing on paying creators based on how their work is used.

What's going on

Some policymakers and industry groups have floated the idea of a “revenue-based charge” for AI, meaning AI companies would pay a fee based on how much money they make, with the goal of protecting creators’ incomes and giving companies clearer legal rules. But the available reporting does not point to a specific, established policy that does this in a direct, proven way.

Instead, the conversation is moving toward systems that look more like a utility meter. In music, some AI music platforms are moving away from one-time licensing fees and toward recurring royalties that are triggered each time a song is played, remixed, or otherwise reused. These payments depend on “attribution” systems, which are ways to keep track of who made what, using tools like metadata (labels attached to files), watermarking (a hidden marker), and audio fingerprinting (like Shazam-style matching).

For journalism, one idea getting attention is statutory licensing. This is a government-set rule that requires payment when news content is used to train AI, similar to a standard fee schedule so publishers do not have to negotiate deal by deal.

What to watch

Legal uncertainty is still a big factor. US court cases are testing whether training AI on copyrighted work counts as “fair use” (a legal exception that can allow reuse without permission), with key rulings expected in 2026. Other countries are also tightening rules, including requirements to label AI content and to use centralized royalty systems, which could raise costs for global AI firms.

Source: Financial Times