This will likely be old news by the time we post this but Anthropic has announced that they will be adding machine-readable watermarking in all AI-generated content.
This is partially because they signed an EU AI Act that requires AI content to be marked. Specifically, it is EU AI Act Article 50(2) - Code of Practice on Transparency of AI-Generated Content.
For Claude users, this means that any model released after Aug 2, 2026 will support marking of all content. “Machine-readable” marking implies that it will not be human-readable but I might be off on that…
They are also updating Claude models that were released before Aug 2 so that they can also support marking. If you use Claude, it will be a safe bet that any model will support some kind of marking.
Anthropic will also provide detection to enable users and third parties to detect Claude’s embedded watermarks.
Before you ditch Claude and run to another provider’s products (like a lot of people have…), keep in mind that any provider that does business in the EU has likely signed the same Act and also marks their output (even if they haven’t announced it)