SpaceXSI has announced that its Grok chatbot will begin embedding invisible watermarks into AI-generated text starting November 7, a step the company says is intended to comply with the European Union's AI Act.
November 7 · SpaceXSI
Grok to Embed Invisible Watermarks in AI-Generated Text
SpaceXSI says its Grok chatbot will begin embedding invisible watermarks into text it generates, a step intended to comply with the European Union's AI Act. Key technical details have not yet been confirmed.
Nov 7
Announced start date for watermarking
Text
Scope includes text, not only images or video
Invisible
Marks are not visible to readers of the output
How the rollout links to EU transparency rules
STEP 1
EU AI Act
Transparency duties for generative AI
→
STEP 2
Watermarking
Invisible marks added to Grok text
→
STEP 3
Detection
Machine-readable identification of origin
What is stated
Watermarking starts November 7
Applies to text output from Grok
Marks are invisible to ordinary readers
Aimed at EU AI Act compliance
Still unclear
How the watermarks are constructed
Whether they can be removed
Whether third parties can verify them
Whether all output or only EU users are covered
Why it matters
Text watermarks are harder to build than image or audio marks, and researchers note they can be weakened or stripped.
Must survive copying, paraphrasing and editing
Must stay invisible to ordinary readers
Limited weight as proof of origin
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article...