Anthropic signed the EU AI Act's Code of Practice and explained how it will apply the rules.

New models launched in the EU after August 2, 2026, will have watermarks. This includes all Claude models. The watermarks will be embedded in text and included in files as metadata.

The watermarks will appear worldwide, not just in Europe, but some platforms may not support them.

The watermarks will not be visible to most people. For text, the watermark will be part of the text and will move with it when copied. For images, Anthropic will use an open standard. More details will be shared later.

Existing models will also get watermarking support, says Anthropic.

The watermarks are not foolproof. Edited or translated text may not have a detectable watermark, and file metadata can be removed.

Anthropic is not the first AI company to make this commitment. Others like OpenAI, Google, and Meta have also done so.

To get more tech news, sign up for Mashable's newsletter or push alerts. You can get updates daily or straight to your device.

The topic is Artificial Intelligence.