EU and UK move toward new rules for AI training on copyrighted work

In short: Governments, led by the European Union and the United Kingdom, are working on new rules to protect creators when their copyrighted work is used to train AI models.

What's going on

The European Parliament has backed a set of proposals that would tighten how AI companies can use copyrighted books, images, music, and other work for training. Training data is the material an AI system learns from, like textbooks used to teach a student.

In March 2026, Parliament passed a resolution calling for clearer copyright rules for generative AI (AI that can produce new text, images, or audio). It also suggested a simple fee, described as 5 to 7 percent of global turnover, to help compensate creators. Another big idea is making opt-outs easier, so creators can clearly say their work cannot be used for AI training, possibly through a central register run by the EUIPO, the EU’s intellectual property office.

The EU is also pushing for more transparency. Lawmakers want AI providers to share more detail about what copyrighted works were used in training and how data was collected. Some related requirements are already in the EU AI Act, including publishing a summary of training data and following opt-out rules, with parts starting to apply in 2025.

The UK is taking a different step, with an ongoing consultation on copyright and AI. It is weighing options like stronger licensing, and debates continue over whether creators should have to opt in or opt out.

What to watch

Countries outside Europe are also exploring approaches. The US is still studying the issue through the Copyright Office, and India has floated a “one license, one payment” idea with centralized royalties. For everyday people, the key question is whether these rules lead to clearer permissions and payments, and whether they change what AI tools can legally learn from.

Source: NYTimes