Anthropic Adopts EU AI Act Standards for Claude Model Watermarking

On Tuesday, August 16, 2026, artificial intelligence safety and research company Anthropic officially announced its commitment to complying with the European Union Artificial Intelligence Act's Code of Practice. As part of this major regulatory alignment, the organization has formally become a signatory to the EU code governing the transparency of AI-generated content across the digital landscape.

The European regulatory framework officially took effect earlier this month on August 2. Under the strict requirements of this newly implemented EU code, artificial intelligence service providers are legally and operationally mandated to transparently and clearly label all material produced by automated systems.

To meet these stringent compliance standards, Anthropic has confirmed that all newly released Claude models launched on or after the August 2 deadline are now equipped with advanced digital watermarking capabilities. Specifically, this system includes specialized machine-readable watermarking for text outputs alongside Coalition for Content Provenance and Authenticity (C2PA) open standard provenance metadata specifically designed for images.

Looking beyond immediate compliance within European markets, Anthropic has outlined plans to apply these protective markings on a global scale. Furthermore, the company intends to extend comprehensive support for these watermarking and metadata features to older legacy Claude models that were originally released prior to the August regulatory deadline.

The broader technology sector is currently adapting to these sweeping transparency rules, with numerous industry players navigating the evolving regulatory environment. While Anthropic moves forward with its implementation of C2PA standards and machine-readable text watermarks, the ongoing enforcement of the EU AI Act continues to shape how automated content generation is managed worldwide.