Anthropic Adds Watermarks In Claude Content As Deepfakes Run Amok
As AI-generated content floods the internet, differentiating machine output from human work has emerged as a defining problem. While regulators…
In response to the rise of deepfake content, major AI developers are taking steps to label AI-generated material. One such company, Anthropic, has announced that it will make watermarking a default feature across all Claude products, including APIs, Claude Code, and enterprise deployments, globally. Claude models launched in the EU on or after August 2, 2026, will carry machine-readable marks, while older models will receive support during a transition period.
Anthropic aims to use both embedded text watermarks and signed provenance metadata for files, with the latter following the C2PA open standard for supported file types. However, the effectiveness of these watermarks is limited, as they are merely a probabilistic signal rather than a definitive proof of Claude's involvement. Critics argue that the Claude approach lacks technical specifications, interoperability with other watermarking systems, and third-party detection access, making it vulnerable to deepfake-enabled fraud and misinformation.
Written by urgent.news from Inc42's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
