conv.

All stories
AIActive · 27h ↳ Anthropic Adds Invisible Watermarks to Claude-Generated Text

Anthropic Adds Invisible Watermarks to Claude-Generated Text

New Claude models silently mark AI-generated text and files to make machine authorship detectable, driven by EU AI Act rules.

3 items21k engagement799 comments TechmemeX
Anthropic Adds Invisible Watermarks to Claude-Generated Text
x.com

Conversation activity · last 28 hours peak 1/30m

Peak 1 item in one 30m at Aug 10, 5 PM; 3 items over 28 hours Aug 10, 5:06 PM — no itemsAug 10, 5:36 PM — 1 item · X 1Aug 10, 6:06 PM — no itemsAug 10, 6:36 PM — no itemsAug 10, 7:06 PM — no itemsAug 10, 7:36 PM — no itemsAug 10, 8:06 PM — no itemsAug 10, 8:36 PM — no itemsAug 10, 9:06 PM — no itemsAug 10, 9:36 PM — no itemsAug 10, 10:06 PM — no itemsAug 10, 10:36 PM — no itemsAug 10, 11:06 PM — no itemsAug 10, 11:36 PM — no itemsAug 11, 12:06 AM — no itemsAug 11, 12:36 AM — 2 items · Techmeme 2Aug 11, 1:06 AM — no itemsAug 11, 1:36 AM — no itemsAug 11, 2:06 AM — no itemsAug 11, 2:36 AM — no itemsAug 11, 3:06 AM — no itemsAug 11, 3:36 AM — no itemsAug 11, 4:06 AM — no itemsAug 11, 4:36 AM — no itemsAug 11, 5:06 AM — no itemsAug 11, 5:36 AM — no itemsAug 11, 6:06 AM — no itemsAug 11, 6:36 AM — no itemsAug 11, 7:06 AM — no itemsAug 11, 7:36 AM — no itemsAug 11, 8:06 AM — no itemsAug 11, 8:36 AM — no itemsAug 11, 9:06 AM — no itemsAug 11, 9:36 AM — no itemsAug 11, 10:06 AM — no itemsAug 11, 10:36 AM — no itemsAug 11, 11:06 AM — no itemsAug 11, 11:36 AM — no itemsAug 11, 12:06 PM — no itemsAug 11, 12:36 PM — no itemsAug 11, 1:06 PM — no itemsAug 11, 1:36 PM — no itemsAug 11, 2:06 PM — no itemsAug 11, 2:36 PM — no itemsAug 11, 3:06 PM — no itemsAug 11, 3:36 PM — no itemsAug 11, 4:06 PM — no itemsAug 11, 4:36 PM — no itemsAug 11, 5:06 PM — no itemsAug 11, 5:36 PM — no itemsAug 11, 6:06 PM — no itemsAug 11, 6:36 PM — no itemsAug 11, 7:06 PM — no itemsAug 11, 7:36 PM — no itemsAug 11, 8:06 PM — no itemsAug 11, 8:36 PM — no items 1 item · 5:36 PM
Aug 118 AM4 PMnow · 9:06 PM

Summary, timeline and people extracted by Claude from 3 items across 2 sources · 1h ago. Quotes are verbatim.

Anthropic confirmed that all Claude models launched from August 2, 2026 onward embed an imperceptible watermark in generated text, and use the C2PA standard for images, so AI-generated content can be flagged after the fact. The move responds to the EU AI Act's Code of Practice on Transparency of AI-Generated Content, whose core transparency obligations began enforcement on August 2, 2026, with penalties up to €35 million or 7% of global turnover for non-compliance.

  • All Claude models launched from August 2, 2026 onward embed an imperceptible watermark in generated text and apply the C2PA standard to image files.
  • The change is a response to the EU AI Act's Code of Practice on Transparency of AI-Generated Content, whose core obligations began enforcement August 2, 2026.
  • Anthropic says the watermark isn't fully conclusive — it can indicate AI involvement but not who was responsible, and can be diluted by heavy editing or very short text.
  • Google, Meta, OpenAI and others have also committed to complying with the EU rules, so switching AI models won't avoid watermarking; the change also raises privacy concerns since watermarks travel with content worldwide.

How it unfolded

  1. Report How-To Geek details watermarking mechanics and limits

    Coverage explains that the watermark isn't fully conclusive, can be diluted by heavy editing or very short text, and raises privacy concerns even though it doesn't directly identify users.

  2. Reaction Polymarket tweet spreads news of Claude watermarking

    A widely shared post from Polymarket's X account announces that Claude will invisibly watermark AI-generated text so it can be detected after being copied and pasted, drawing over 21,000 reactions.

    “JUST IN: Claude will now invisibly watermark AI-generated text so it can be detected after being copied & pasted.”

    @Polymarket · X ↗
  3. 9 days quiet
  4. Event EU AI Act core transparency obligations take effect

    The EU begins enforcing core obligations around transparency and high-risk AI systems under its AI Act, prompting companies to comply with watermarking requirements for generated content.

  5. Event Claude models begin watermarking generated content

    Anthropic's support page confirms all Claude models launched from this date add an imperceptible watermark to generated text and use C2PA for image files like JPEG and PNG.

What people are saying verbatim

“JUST IN: Claude will now invisibly watermark AI-generated text so it can be detected after being copied & pasted.”

@Polymarket, X account · x ↗ · Aug 9

“imperceptible”

Anthropic, AI developer · Anthropic support page, via How-To Geek · Aug 10

“Anthropic notes that the watermarking isn't "fully conclusive."”

How-To Geek, Technology publication · How-To Geek ↗ · Aug 10