Anthropic to implement machine-readable marking across new AI models to ensure content transparency

San Francisco (California) [US], August 12 (ANI): Claude's new models will mark AI-generated content from day one as part of compliance commitments under the European Union AI Act's Article 50(2) Code of Practice on Transparency of AI-Generated Content.
According to the company, the commitment applies to Claude models launched in the EU on or after August 2, 2026, which will support machine-readable marking at launch.
To execute this, two complementary techniques mark the generated and processed content: watermarks embedded directly into text and signed provenance metadata attached to files. The embedded text watermark functions invisibly without altering quality or readability, traveling with the text when copied. For supported file formats such as .svg, .png, or .jpg, signed metadata follows the Coalition for Content Provenance and Authenticity (C2PA) open standard.
As per the company, the transparency measures apply across various platforms and global regions where the models operate.
"Marking works everywhere you use Claude. Marks will apply to output from supported Claude models across Claude Platform (API), Claude, Claude Code, Claude Cowork, and Claude Tag, and wherever Claude is offered, worldwide. Some platforms or features may not support certain marking types," the company stated.
The organization noted that a transition period remains underway for earlier model iterations launched prior to the specified deadline.
"The law includes a transition period for Anthropic models launched before August 2, 2026, and we're working to add marking support for those models as well," it noted.
Cloud distribution channels will also carry specific marking mechanisms based on platform capabilities.
"Embedded watermarks will apply when supported Claude models are accessed through AWS, Google Cloud, or Microsoft Foundry. Signed provenance metadata may not be supported on every platform, depending on the features each platform offers," the company detailed.
Additionally, technical mechanisms are being developed to assist users and third parties in detecting these embedded marks and metadata labels. However, the documentation highlights inherent technical limitations, noting that detecting a mark indicates content processing by the system but does not conclusively confirm full authorship or original provenance.
"Claude may not be the original author. People often use Claude to proofread, translate, summarize, or convert files. The output can carry a Claude mark even if the underlying ideas, text, or data originated from another source," the company clarified.
It added that the absence of a mark does not guarantee content was unassisted by artificial intelligence, as short passages, heavy editing, or metadata stripping during re-saving can remove detectable signals.
Third-party developers deploying the system are advised to independently assess legal compliance obligations. (ANI)