The European Commission has proposed draft changes aimed at making generative AI systems more transparent. Under the plan, providers would need to publish summaries of the data used to train their models and add visible watermarking for AI-generated content.

The measures are designed to give users clearer information about how these systems are built and when content has been produced by artificial intelligence. Officials said the draft rules would apply to providers of generative AI models after adoption, with a 12-month window to comply.

The proposal adds to the EU’s broader effort to set guardrails for fast-moving AI tools while trying to balance innovation with public accountability. Supporters say the changes could help curb misinformation, reduce confusion around synthetic content, and improve oversight of powerful AI products.

The draft still needs to move through the EU’s legislative process before it becomes binding. If approved, it could set a precedent for other regulators weighing similar transparency requirements for AI developers.