Anthropic Adds Machine-Readable Watermarks to Claude Text Under EU AI Act Rules Anthropic will begin embedding imperceptible machine-readable watermarks into text generated by its new Claude models starting August 2, 2026. This move aligns with transparency requirements under Article 50 of the EU AI Act, which mandates greater accountability for AI systems. The company has also signed the European Union’s Code of Practice for AI-generated content, reinforcing its commitment to compliance. The new watermarking system will apply globally, not just to European users, and will be integrated into all outputs from Claude models launched after August 2, 2026. The watermarks will be applied at the model level, ensuring they persist across all Claude products and platforms. For text, the system will use invisible signals that remain embedded even after users copy and paste responses. Some marks may survive editing, though Anthropic acknowledges that heavy modifications could reduce their detectability. The watermarking process will extend to supported files, such as images, through a separate system based on the Coalition for Content Provenance and Authenticity (C2PA) standard. Files will carry signed metadata to verify their origin and detect unauthorized alterations. Anthropic’s approach aims to provide platforms and organizations with a technical signal to identify AI-generated content. The company plans to publish detailed technical documentation to help users and third parties verify the presence of watermarks. This includes tools for detecting the marks, which could aid in tracking AI-generated material across workflows. However, Anthropic cautions that neither text watermarks nor file metadata should be considered definitive proof of authorship.#anthropic #claire #eu_ai_act #c2pa #code_of_practice_ai_generated_content
Anthropic Announces AI Text Watermarking Compliance with EU Regulations Anthropic, a leading AI model developer, has announced it will implement watermarking for text generated by its AI systems, including the Claude series, to align with European Union regulations. The company confirmed this policy update in a revised support page, emphasizing its commitment to adhering to the EU AI Act’s Transparency Code, which became enforceable on August 2, 2026. This regulation mandates that AI-generated or edited content be marked in a way that allows other systems to identify its origin. Under the new policy, all AI models released after August 2 will automatically include technology to watermark both textual and file-based outputs. For files, Anthropic will utilize the C2PA (Content Credentials for the Public Web) open standard, a framework designed to verify the authenticity and origin of digital content. The company also stated it will provide support for older models, ensuring the watermark remains intact even when users copy and paste the text. According to the support page, the watermark is embedded as part of the text itself, meaning it will persist through copying, pasting, and some editing processes. The watermark will be applied at the model level, ensuring its presence regardless of which Claude product or platform the text originates from. The implementation of watermarking extends to various Anthropic products, including the Claude platform API, Claude, Claude Code, Claude Cowork, and Claude Tag. This measure aims to enhance transparency and accountability in AI-generated content, particularly in contexts where distinguishing human-created text from AI-generated material is critical.#anthropic #claude #eu_regulations #eu_ai_act #c2pa
