According to TechCrunch, Anthropic has announced that it will embed watermarks in texts generated by its models (including Claude) to comply with the transparency requirements of the EU's Artificial Intelligence Act. The Act, which came into effect on August 2, stipulates that AI companies must add machine-readable identifiers to generated or edited content to enhance transparency. Anthropic stated that the watermarks will be embedded in the text by default, without affecting the reading experience, and can still be detected after copying, pasting, or partial modifications. Additionally, files such as images generated by Claude will also include digital provenance metadata compliant with the international C2PA standard to record generation or processing information and whether they have been tampered with. This measure applies uniformly across the Claude web version, API, and partner platforms such as AWS and Google Cloud, and is enabled globally.