Now Reading: Anthropic watermarks Claude AI text to comply with EU rules
-
01
Anthropic watermarks Claude AI text to comply with EU rules
Anthropic has signed the EU AI Act’s Code of Practice on Transparency of AI-Generated Content. Starting August 2, all Claude models launched in the European Union will carry machine-readable watermarks in text,…
The Story in Brief
Anthropic has signed the EU AI Act’s Code of Practice on Transparency of AI-Generated Content. Starting August 2, all Claude models launched in the European Union will carry machine-readable watermarks in text, images, and other files. The invisible watermark is woven into text and survives copying, pasting, and light editing, the company says.
The mark applies globally across platforms including AWS, Google Cloud, and Microsoft Foundry. However, Anthropic warns that false positives and negatives are possible: heavily edited or very short text may lose the watermark, and human content processed by AI could also carry the mark. Tools to help users detect these watermarks will be released soon.
The Indian Opinion
Some will claim this solves the AI detection problem or, conversely, that watermarking is useless because it can be stripped. Both miss the point. The watermark is a transparency measure, not a silver bullet. It is honest about limitations, false positives, false negatives, and applies even to proofreading, a stricter stance than the EU requires. The real test is whether Anthropic’s detection tools work reliably in practice and whether rivals adopt similar marks. Watch for the first independent audits of the watermark’s resilience against simple rewording attacks.
Sources (2): thehindu.com, gsmarena.com
This story was synthesised by AI from the 2 sources linked above.
Updated: this story now draws on 2 sources.
