EU AI Act

The EU AI Act is the European Union's law governing artificial intelligence. Think of it like an ingredients label on food or a safety certification mark on electronics — it's a set of rules requiring AI-generated outputs and AI services to carry labels like "this was made by AI" or "this carries certain risks," along with related management obligations.

There's a real example of this law changing how companies behave. Anthropic signed onto the law's transparency provisions and began embedding invisible markers in text generated by its chatbot Claude, as well as attaching provenance information to image files. Technically, Anthropic only needed to apply this to products released in the EU, but it rolled the feature out uniformly to its products worldwide, not just in the EU.

In other words, while the EU AI Act only has legal force within Europe, major AI companies often apply a single policy identically everywhere in the world, so its practical influence extends well beyond Europe's borders.

You'll see it in sentences like: "After signing onto the EU AI Act's transparency code, Anthropic began building labeling features into new Claude models by default." This means the EU AI Act itself doesn't specify watermarking technology — rather, companies chose watermarking as their method of complying with the regulatory requirement to make AI-generated content identifiable. The key is not confusing the law itself with the technical implementation companies choose.