Now Reading: Anthropic to watermark Claude AI text content globally
-
01
Anthropic to watermark Claude AI text content globally
Anthropic, backed by Google and Amazon, will add invisible watermarks to text generated by its Claude AI, complying with the European Union's AI Act. The mark is embedded at the model level,…
The Story in Brief
Anthropic, backed by Google and Amazon, will add invisible watermarks to text generated by its Claude AI, complying with the European Union's AI Act. The mark is embedded at the model level, travels with copied text, and may survive some editing. The Hindu reports that even light actions like proofreading or translation using Claude could carry the mark. Critics warn this could wrongly label human work as AI-generated. Anthropic cautions that false positives and negatives are possible, especially with short texts or format changes. The full technical details are not yet public.
The Indian Opinion
The panic around Anthropic's invisible watermark misses a key point: detection is not foolproof. Both false positives and false negatives are acknowledged by the company itself. The narrative that this will end undisclosed AI use is overblown, just as earlier AI detectors led to unfair accusations. Similarly, the fear that any light editing with Claude will taint human work ignores that watermarking remains experimental. The real test: can third-party tools reliably identify the watermark without wrongly flagging human text?
Source: thehindu.com
This story was synthesised by AI from the source linked above.
