10/20/2026
Are the days of posting AI content without anyone knowing coming to an end?
Anthropic is adding an invisible watermark to text written by new Claude models.
You won't see the watermark. It is created through patterns in the words Claude chooses, so copying the answer into Word, an email or a website won't remove it. Some editing may leave the mark intact too.
Anthropic plans to provide tools that can estimate whether Claude helped write a piece of text.
Those tools cannot identify the person who used Claude, show what they asked or tell how much of the finished document was written by them.
A document may carry the watermark even if Claude was only used to rewrite part of it.
Businesses can keep using Claude for drafts and editing, but staff should know that its involvement may be detectable. Anything sent to a client or published under the company name still needs to be checked for accuracy.
If your team is already using AI at work and you're not sure what rules you need, let's have a chat.
https://emtcc.net/contact-us/
