3 weeks ago
Anthropic Adds Invisible Watermarks to Claude Text for EU Rules
A company called Anthropic makes a very smart computer helper named Claude that can write and answer questions.
Now, Anthropic says it will add a secret, invisible stamp to text that Claude helps with.
People cannot see this stamp, and it stays even when the text is copied and pasted somewhere else.
The stamp shows that Claude helped with the text, but it does not mean Claude had the ideas by itself.
Sometimes people use Claude to fix spelling, translate, or shorten their own writing.
The stamp is there to follow new rules from the European Union, which wants machine-made content to be marked.
The new stamps will be on Claude models released after August 2, and older models will also be updated.
Anthropic will use the stamp everywhere in the world, not only in Europe.
Many big companies signed a promise to follow these rules, and Google built its own secret stamp called SynthID.
Anthropic announced it will add an invisible watermark to text processed by its Claude chatbot to comply with European Union regulations.
The watermark indicates text was processed by Claude, but does not confirm Claude generated it.
The mark is invisible to humans and remains even after text is cut and pasted, but very short or heavily edited text may not carry it.
Claude models launched on or after August 2 will feature the watermarks, with earlier models to be upgraded.
Anthropic will apply the watermarks worldwide, not just in Europe, where AI-generated content must feature machine-readable markings.
- Who
- Anthropic, the American AI company behind the Claude chatbot.
- What
- Anthropic announced it will add invisible watermarks to text processed by Claude, with signed provenance metadata for supported file types like .jpg.
- Where
- Announced in San Francisco; the watermarking will apply worldwide, not just in Europe.
- When
- The announcement follows new EU AI regulations; watermarks will appear on Claude models launched on or after August 2.
- Why
- To comply with European Union regulations requiring AI-generated content to feature machine-readable markings, with stiff fines for violations.
Key facts
- Company
- Anthropic
- Product
- Claude chatbot
- Watermark visibility
- Invisible to humans; survives cut-and-paste
- Effective models
- Claude models launched on or after August 2; earlier models will be upgraded
- Geographic scope
- Worldwide
- Reason
- Compliance with EU AI regulations requiring machine-readable markings on AI content
- Industry context
- About 190 organizations signed a voluntary EU code of good practice, including Google, Meta, Microsoft, Mistral and OpenAI
- Related technology
- Google's SynthID, a non-public advanced digital watermarking and detection technology
Quotes
Anthropic representative
Company spokesperson
“The watermark means the text was processed by Claude but does not confirm the source”
deccanchronicle.com
Sources
Related news
Anthropic adds invisible watermarks to Claude AI-generated text
Anthropic Adds Invisible Watermark to Claude Text Worldwide
Developers Race to Remove Anthropic’s Invisible AI Watermarks
Anthropic adds invisible watermark to Claude AI-generated text
Anthropic Explains New Watermarking for Claude and Its Limits
Cardano founder launches free tool to remove Anthropic's AI watermarks
Social Media Companies Intensify Efforts to Identify Child Users
Yorkville America Expands Beyond Trump-Linked ETFs With AI Strategy
Meta’s $18-Billion Child Safety Settlement Faces Praise and Criticism
School Awards Honour Courage, Commitment and Educational Excellence
AI Expansion Puts Jamnagar and Thane Under Water Stress
