Apfelpatient
  • Home
  • News
  • Rumors
  • Tips & Tricks
  • Reviews
  • Insights
No result
View All Results
  • iPhone 18
  • iPhone 18 Pro
  • iPhone Duo
  • Apple Home
  • Siri AI
  • HomePod
  • HomePod mini
  • HomePod mini 2
  • MacBook Pro
  • Apple TV 4K
  • Apple Watch Series 12
  • Apple Watch Ultra 4
  • AirPods 5
  • iOS 27
  • iPadOS 27
  • macOS
  • MacBook Neo
  • iPhone 17e
  • AirTags
  • iOS 26
  • iPhone 17
  • iPadOS
  • iMac
  • Mac mini
  • Apple TV
  • iPad Pro
  • iPhone Air
  • Apple Vision Pro
  • iPhone 16
  • AirPods 4
  • Apple Watch Ultra
  • MacBook Air
  • iPad
  • AirPods Max
  • AirPods Pro 3
  • Mac Studio
  • Studio Display
  • iPad Air
  • iPad mini
  • watchOS 27
  • Apple CarPlay
  • Apple Pay
  • watchOS 26
  • Home
  • News
  • Rumors
  • Tips & Tricks
  • Reviews
  • Insights
No result
View All Results
Apfelpatient
No result
View All Results
  • iPhone 18
  • iPhone 18 Pro
  • iPhone Duo
  • Apple Home
  • Siri AI
  • HomePod
  • HomePod mini
  • HomePod mini 2
  • MacBook Pro
  • Apple TV 4K
  • Apple Watch Series 12
  • Apple Watch Ultra 4
  • AirPods 5
  • iOS 27
  • iPadOS 27
  • macOS
  • MacBook Neo
  • iPhone 17e
  • AirTags
  • iOS 26
  • iPhone 17
  • iPadOS
  • iMac
  • Mac mini
  • Apple TV
  • iPad Pro
  • iPhone Air
  • Apple Vision Pro
  • iPhone 16
  • AirPods 4
  • Apple Watch Ultra
  • MacBook Air
  • iPad
  • AirPods Max
  • AirPods Pro 3
  • Mac Studio
  • Studio Display
  • iPad Air
  • iPad mini
  • watchOS 27
  • Apple CarPlay
  • Apple Pay
  • watchOS 26

ChatGPT: Invisible watermark for texts in the EU

Milan Jovicic by Milan Jovicic
October 5, 2026 - 23:42 CEST
in Apple News
0
White OpenAI logo on a blue gradient – ChatGPT watermark for AI-generated text in the EU

Image: OpenAI / Apfelpatient

WhatsAppFacebookEmail
Threads

This article may contain affiliate links. Learn more on our Partner Program page.

Make Apfelpatient a preferred source One click – and you'll see us more often on Google

Texts from ChatGPT and Codex will soon receive a tag in the EU that is invisible to everyone but technically detectable. OpenAI is thus responding to the labeling requirement of the AI Act – while simultaneously acknowledging how easily the signal can be weakened.

OpenAI, the AI research and practical application company, has outlined in a statement on EU rules for text origin how it will label AI-generated texts in the future. This is in response to Article 50 of the AI Act, whose transparency obligations have been in effect since August 2, 2026.

Afterwards, providers of generative AI must mark their output in a machine-readable way so that it is recognizable as artificially generated. For text, OpenAI uses a watermark embedded in the word choice itself. This will apply to all ChatGPT plans, including the free plan.

Key Takeaways

  • In the coming weeks, texts from ChatGPT and Codex in the EU will receive an invisible watermark, across all tariffs.
  • Outside the EU, ChatGPT remains without a watermark; this also applies to Switzerland.
  • Customers worldwide can now voluntarily activate a watermark in the API; it remains off by default.
  • Initially, only selected researchers and professional organizations will receive the detector that recognizes the watermark.

This is how textGrain works

The method is called textGrain. It inserts an invisible statistical signal into the model's word choice: where several formulations fit equally well, the selection follows a hidden pattern. A detector searches a text for precisely this pattern.

According to OpenAI, this has no significant impact on the quality of the answers. In eight benchmarks of the current Astra model, the results with and without watermarking are very close, sometimes slightly higher, sometimes slightly lower. OpenAI also intends to release the technology as open source; a technical report already describes it.

How reliable is the detection?

OpenAI outlines the limitations of its method, both internally and numerically. With a false positive rate of one percent, the detector identified the watermark in approximately 80 percent of passages containing 200 tokens and in approximately 95 percent of passages containing 400 tokens, based on texts related to psychology. In mathematics, where word choice allows for less leeway, the figures were significantly lower.

Post-processing has an even stronger effect:

Replaced words (400 tokens)Recognition rate
noneapproximately 92 %
10 % by synonymsapproximately 66 %
25 % by synonymsapproximately 17 %

Furthermore, a watermark says nothing about how much a person contributed to the text, who owns it, who created it, or whether it is accurate. Conversely, its absence does not prove that a person wrote the text.

Not the first provider

With this move, OpenAI is following a similar approach to Anthropic. The provider of Claude had announced in the summer that it would mark texts using slightly modified word choices.

This does not create a shared detector. OpenAI's tool only checks whether a text contains an OpenAI watermark – it does not recognize texts from other models as AI-generated.

A short-range signal

I consider the watermark in its current form primarily a legal obligation, not a tool for reliably detecting AI-generated texts in everyday life. Replacing a quarter of the words is enough to reduce the detection rate to less than a fifth, and the detector isn't public anyway.

For users in Germany and Austria, nothing visibly changes in ChatGPT; the watermark is solely due to the wording. In Switzerland, which is not part of the EU, texts remain without a watermark.

Do you think a watermark in AI-generated texts makes sense if it can be circumvented simply by changing a few words? Let us know in the comments where you would draw the line.

Was this article helpful?
YesNo
Tags: ChatGPTEUOpenAI
SendShareSend
Share

Our Amazon Storefront

A handpicked selection of products for iPhone, Mac and more – sorted by topic and updated regularly.

Shop Now

This post contains affiliate links (including Amazon). We earn a small commission on qualifying purchases – at no extra cost to you. Learn more on our Partner Program page.

Previous Post

Apple Music: New logo appears in iOS 27.2

Next Post

macOS 27.2: Code prepares touch gestures for the Mac

Milan Jovicic

Milan Jovicic

Milan founded Apfelpatient in 2016 and has been responsible for all editorial content since 2018 – news, rumors, how-to guides and product reviews. Here, Apple devices aren't review units kept for two weeks, they're everyday tools: from iPhone to MacBook Pro, MacBook Air and iMac all the way to Apple Vision Pro, at least one device from every product category is in daily use, many of them replaced every year. Every menu path in a guide is checked on the device before it goes live.

Related posts

Presumed Innocent: Season 2 Arrives on February 19

by Milan Jovicic
November 20, 2026 - 17:39 CET
Presumed Innocent season 2: Rachel Brosnahan as defense attorney Lelia Reynolds in the courtroom

Presumed Innocent season 2 premieres February 19 on Apple TV, with Rachel Brosnahan and Matthew Rhys as a married couple in a murder trial.

Read moreDetails

Apple is working on an AI agent for apps, browsers, and code

by Milan Jovicic
November 20, 2026 - 01:18 CET
Apple AI agent: Siri AI icon with a shimmering sphere on a white tile against a dark gray gradient

Apple is working on an AI agent that will work with apps, browsers, and code. This is revealed in a job posting that also mentions on-device models.

Read moreDetails

Apple vs. Epic: Apple says Epic's search terms are too broad

by Milan Jovicic
November 19, 2026 - 10:13 PM CET
Apple vs. Epic: illustrated file folder showing case number 4:20-cv-05640 of the commission proceedings on a dark blue background

Apple vs. Epic: Apple says Epic's search terms are too broad and offers a sample to the court instead of all 122,000 documents.

Read moreDetails

Amazon Black Friday Week: The deals for Apple users

by Milan Jovicic
November 19, 2026 - 17:16 CET
Amazon Black Friday week: iPhone Duo showing the words Black Friday Deals on its display

Amazon Black Friday Week until November 30th: The offers for Apple users and which of these prices are genuine Black Friday deals.

Read moreDetails

iCloud backdoor: Court confirms British order

by Milan Jovicic
November 19, 2026 - 3:59 PM CET
iCloud backdoor: Unfolded iPhone Duo in a folio case showing the Advanced Data Protection settings page with the feature turned on

A British court overturns the secrecy: The order to Apple exists, the lawsuit against the iCloud backdoor becomes public.

Read moreDetails
Load More

Categories

  • Apple Insights
  • Apple Rumors
  • Apple News
  • Apple Tips & Tricks
  • iPhone news, rumors and tips
  • Mac and MacBook News
  • Reviews
Make Apfelpatient a preferred source One click – and you'll see us more often on Google

About APFELPATIENT

APFELPATIENT brings you the latest Apple news, product updates, guides, reviews and tips across the entire Apple ecosystem — from the iPhone to the Mac to the Apple Vision Pro. From the first rumors to confirmed news: researched responsibly.

Follow Apfelpatient

Facebook Instagram YouTube Threads Threads

Company

  • About Apfelpatient
  • Contact
  • Author Profiles

Community

  • Netiquette
  • Push notifications
  • RSS feed

Legal

  • Legal Notice
  • Privacy Policy
  • Terms of Use
  • Cookie settings
  • Affiliate Program

Resources

  • Sitemap

© 2026 Apfelpatient. All rights reserved.

No result
View All Results
  • Home
  • News
  • Rumors
  • Tips & Tricks
  • Reviews
  • Insights

© 2026 Apfelpatient. All rights reserved. Site map

Change language to Deutsch