iStockphoto / Robert Way/Alexeg84
Young woman writing travel notes in diary during sunrise with beautiful sun light and mountain landscape on the background, close-up.
Corporate America and schools across the globe are about to receive a huge shake-up due to a new policy from the European Union. The EU’s Code of Practice on Transparency of AI-generated Content has forced Claude and other Artificial Intelligence models to embed digital watermarks in AI-generated text in order for any text written by AI to be detectable as such.
In an update to its Support page yesterday, Anthropic’s Claude published guidelines on how the digital watermark will work and when it will be enacted. This is certain to have a seismic impact across the world as people have been passing off AI slop as their own work and taking credit for it.
Anthropic To Include ‘Imperceptible’ Digital Watermark On AI-Written Text From Claude
There has been a meme floating around for months where someone claims they heard someone in Gen Z / Alpha claim nobody could ever write a 10-page book report without the use of artificial intelligence and the response is how we used to do it without ever having read the book. It seems that those people claiming they could never write a paper without AI, or an email, book, thesis, presentation, etc., are in for a rude awakening.
In Anthropic’s announcement, they wrote “new models will mark AI-generated content from day one. Claude models launched in the EU on or after August 2, 2026 will support machine-readable marking at launch. Generated text will carry embedded watermarks, and generated files will include digitally signed provenance metadata where supported.”
There could be an interesting monetization strategy at play here. Anthropic could theoretically charge institutions (companies, schools, etc) for the service that detects the ‘imperceptible’ digital watermark and at least in the case of high schools and universities that could be a very, very lucrative product.
Anthropic’s announcement made it clear that it wasn’t just Claude but all of the Claude products. They wrote “marking works everywhere you use Claude. Marks will apply to output from supported Claude models across Claude Platform (API), Claude, Claude Code, Claude Cowork, and Claude Tag, and wherever Claude is offered, worldwide. Some platforms or features may not support certain marking types.”
It continued saying “we’ll help you detect Claude’s marks. We’ll support users and other third parties to detect Claude’s marks, as the Code requires, and we’ll share details in forthcoming documentation. Existing models are in progress. The law includes a transition period for Anthropic models launched before August 2, 2026, and we’re working to add marking support for those models as well.”
This is a fantastic development across the board. At the school-age level it will prevent students from passing of AI slop as their own work and have them learning how to use their brains again. Across the corporate world, clients will no longer be paying exorbitant sums of money for reports and products that were built by consultants using AI and charging as if they’d done the work themselves.
How Will It Work?
In its Support page update this week, Anthropic went into detail about how the digital watermark will appear on AI-created content. The beauty is because it cannot be seen by the naked eye that human editing of any AI-written text will likely have little to no impact on the watermark.
They wrote ““When a supported Claude model generates text, it weaves an imperceptible watermark directly into the text itself. You won’t see it, and it doesn’t change the meaning, quality, or readability of Claude’s response. Because the watermark is part of the text, it will travel with the text when it’s copied and pasted elsewhere, and may persist through some editing. Watermarking will be applied at the model level, which means it will be present no matter which Claude product or surface the text comes from.”
This will all be much easier in files such as .jpg and .png where the watermark can/will be embedded in the meta data. An imperceptible digital watermark on AI-written text will be tricky but they seem to have sorted it out.
There are limitations however. Anthropic wrote that “a detected mark provides a signal that content was processed by Claude, but is not fully conclusive. Detecting a Claude mark tells you that the content may have been processed by Claude. It does not, on its own, confirm the full provenance of the content.”
An example they cit here is using AI to proofread or summarize text and/or convert files. This could carry a digital watermark on work that might have been 100% human generated.
They also point out that the lack of a watermark is not proof there was no AI used in the process. Citing Claude AI models that were released prior to the digital watermark process and/or text that has been heavily edited after it was generated using AI as ways to circumvent the watermark.
On a personal note, I love everything about this. I cannot tell you how many times I’ve received comments/emails from a reader complaining about AI use and 10 times out of 10 it is because there was a typo. Each time I respond to the effect of “if this was written by AI don’t you think they wouldn’t have made typos/misspellings? That is just human error and my own fat fingers being slow.” And like clockwork I receive a similar message/email a week later.
We use zero AI here at BroBible. Our team have all been doing this for years if not decades and for better or worse, any mistakes we might make are our own and not AI slop.
I cannot say the same for the other publishers who have popped up in recent years and started pumping out articles that are nearly carbon copies of ours and published less than 10 minutes after our articles go live, or just long enough for whoever re-wrote the article using AI to create an image for the article and hit publish.
Seeing that type of leeching exposed for what it is in the months ahead thanks to the EU AI Act’s Transparency Code is going to be quite cathartic. And for what it’s worth, after I finished writing this article I threw it in an ‘AI checker’ just to see if it would falsely detect my writing as AI and the only portion detected was Anthropic’s announcement, which they presumably wrote using AI.