Anthropic: Invisible watermark for AI texts
190 companies have already signed a voluntary code of conduct, including Google, Meta, Microsoft, Mistral and OpenAI
- Center-left1
- Public / State1
no rewrites detected — all voices distinct
Summary
190 companies have already signed a voluntary code of conduct, including Google, Meta, Microsoft, Mistral and OpenAI. The US company Anthropic has announced that texts from its chatbot Claude will be marked with an invisible watermark in order to comply with the EU requirements for dealing with artificial intelligence (AI). According to the company, the function is invisible to users and is preserved even if a text is copied and reinserted elsewhere.
Furthermore, Older versions will be updated later. August released Claude versions, the watermark is already integrated, according to the company. EU: Label AI-generated content In after 2.
Cross-referenced from 2 sources.
Factual coreconfirmed by several independent voices
190 companies have already signed a voluntary code of conduct, including Google, Meta, Microsoft, Mistral and OpenAI.
reliability moderate2/2 sourcesThe US company Anthropic has announced that texts from its chatbot Claude will be marked with an invisible watermark in order to comply with the EU requirements for dealing with artificial intelligence (AI).
reliability moderate2/2 sourcesAccording to the company, the function is invisible to users and is preserved even if a text is copied and reinserted elsewhere.
reliability moderate2/2 sourcesOlder versions will be updated later.
reliability moderate2/2 sources
Reported detailssecondary facts, each attributed to its source
August released Claude versions, the watermark is already integrated, according to the company.
according to ORF NewsIn Claude versions after the 2.
according to DIE ZEITEU: Label AI-generated content In after 2.
according to ORF NewsAccording to the company, the watermark has already been integrated.
according to DIE ZEIT
Disputedincompatible versions — to verify
No factual contradiction detected between sources.
Framing by sidesame fact, different words — loaded terms highlighted
No notable framing divergence.
Blind spotwhat one side keeps silent
No blind spot detected: every side covers the same facts.