AI news you understand right away.

OpenAI adds an invisible mark to text in the EU. Rewrite a quarter of it, and the mark is almost gone.

OpenAI announced on Monday 5 October that text from ChatGPT and Codex in the EU will get an invisible watermark. The technology is called textGrain. OpenAI says this will help it meet Article 50 of the EU AI Act, the European AI law. That article sets rules on transparency. The rules have applied since 2 August 2026. OpenAI will introduce the watermark over the coming weeks, in the EU only. It applies to all subscriptions, for users who are eligible. The mark stays in place when you copy and paste text. Do you put a property description from ChatGPT into your CRM (your customer system) or onto your website? The mark goes with it. A special program, called a detector, can recognise the pattern. For now, only approved researchers and organisations may use that detector. So you cannot check for the mark yourself, and neither can a buyer or a colleague. The next deadline is 2 December 2026. It applies to systems that were already on the market. textGrain slightly influences the choice of words. There are usually several options for each next word. ChatGPT then chooses words according to a pattern that a detector can find. The pattern is in the words themselves, not in a separate file. That is why it remains after you copy and paste. But the watermark has a weak spot. If someone changes a quarter of the words, OpenAI says detection falls from almost 100% to below 20%. If you rewrite a description a lot, little of the mark remains. If you use the text almost unchanged, the mark stays. A missing watermark does not prove that a person wrote the text. The mark also says nothing about whether the text is accurate. The facts in your listing, such as floor area, year of construction and energy rating, still need your own check. Do you use ChatGPT to write a description? Then you could note in your records that AI was used for the first draft, and which facts you checked yourself.

OpenAI announced on Monday 5 October that text from ChatGPT and Codex in the EU will get an invisible watermark. The technology is called textGrain. OpenAI says this will help it meet Article 50 of the EU AI Act, the European AI law. That article sets rules on transparency. The rules have applied since 2 August 2026. OpenAI will introduce the watermark over the coming weeks, in the EU only. It applies to all subscriptions, for users who are eligible. The mark stays in place when you copy and paste text. Do you put a property description from ChatGPT into your CRM (your customer system) or onto your website? The mark goes with it. A special program, called a detector, can recognise the pattern. For now, only approved researchers and organisations may use that detector. So you cannot check for the mark yourself, and neither can a buyer or a colleague. The next deadline is 2 December 2026. It applies to systems that were already on the market. textGrain slightly influences the choice of words. There are usually several options for each next word. ChatGPT then chooses words according to a pattern that a detector can find. The pattern is in the words themselves, not in a separate file. That is why it remains after you copy and paste. But the watermark has a weak spot. If someone changes a quarter of the words, OpenAI says detection falls from almost 100% to below 20%. If you rewrite a description a lot, little of the mark remains. If you use the text almost unchanged, the mark stays. A missing watermark does not prove that a person wrote the text. The mark also says nothing about whether the text is accurate. The facts in your listing, such as floor area, year of construction and energy rating, still need your own check. Do you use ChatGPT to write a description? Then you could note in your records that AI was used for the first draft, and which facts you checked yourself.

OpenAI gives ChatGPT text in the EU an invisible watermark. Teachers cannot use the detector for now. The watermark is also easy to weaken.

OpenAI announced on Monday 5 October that text from ChatGPT and Codex in the EU will get an invisible watermark. Codex is OpenAI's tool for writing computer code. OpenAI is doing this to comply with Article 50 of the AI Act, the European law on AI. That article is about transparency and has applied since 2 August 2026. The technology is called textGrain. The watermark stays in place when you copy and paste text. The rollout will take the next few weeks and applies only in the EU. Eligible users will get the watermark with all subscriptions. Customers who use OpenAI through an API can switch on the watermark anywhere in the world. An API is a link that lets other software use the model. The watermark is off by default there. The next deadline is 2 December 2026. It applies to systems that were already on the market. For now, only approved researchers and organisations can use the detector. So you cannot read the watermark yourself. textGrain changes the model's choice of words very slightly. A language model chooses from several suitable words each time. If that choice is guided a little, it creates a pattern. You cannot see the pattern. A detector can recognise it. The pattern is in the words themselves. That is why it stays in place after copying and pasting. This is also its weak point. OpenAI says detection falls from almost 100% to below 20% if someone changes a quarter of the words. So a student can almost remove the watermark by rewriting a quarter of the words. A text without a watermark does not prove that a person wrote it. The watermark is only in text from ChatGPT and Codex, and only where it has been introduced. If you ever get access to the detector, a score is a clue, not proof. If you want to see what a student can do on their own, you could have part of the assignment written in class.

OpenAI announced on Monday 5 October that text from ChatGPT and Codex in the EU will get an invisible watermark. Codex is OpenAI's tool for writing computer code. OpenAI is doing this to comply with Article 50 of the AI Act, the European law on AI. That article is about transparency and has applied since 2 August 2026. The technology is called textGrain. The watermark stays in place when you copy and paste text. The rollout will take the next few weeks and applies only in the EU. Eligible users will get the watermark with all subscriptions. Customers who use OpenAI through an API can switch on the watermark anywhere in the world. An API is a link that lets other software use the model. The watermark is off by default there. The next deadline is 2 December 2026. It applies to systems that were already on the market. For now, only approved researchers and organisations can use the detector. So you cannot read the watermark yourself. textGrain changes the model's choice of words very slightly. A language model chooses from several suitable words each time. If that choice is guided a little, it creates a pattern. You cannot see the pattern. A detector can recognise it. The pattern is in the words themselves. That is why it stays in place after copying and pasting. This is also its weak point. OpenAI says detection falls from almost 100% to below 20% if someone changes a quarter of the words. So a student can almost remove the watermark by rewriting a quarter of the words. A text without a watermark does not prove that a person wrote it. The watermark is only in text from ChatGPT and Codex, and only where it has been introduced. If you ever get access to the detector, a score is a clue, not proof. If you want to see what a student can do on their own, you could have part of the assignment written in class.

A hidden marker makes text recognisable, as long as no one changes many words

OpenAI is adding invisible watermarks to text from its AI tools ChatGPT and Codex. This applies to users in the EU. The feature is called textGrain. The technology changes the choice of words slightly. A detector can recognise this pattern. A detector is a program that looks for a marker like this. The marker stays in the text if someone copies it and pastes it somewhere else. Readers cannot see the marker. It is built into a pattern in the text itself, not shown in a message alongside it. The rollout will take place over the coming weeks and applies to all subscriptions. The feature is available to users who are eligible. For HR, this matters for job adverts, internal messages and other text that employees create with AI. The marker makes AI-generated text easier to recognise. It does not prove it for certain. According to OpenAI, recognition falls from almost 100 per cent to less than 20 per cent if someone changes a quarter of the words. For now, the detector is available only to approved researchers and organisations. If the marker is missing, that does not prove a person wrote the text. So the marker alone is not a sound basis for deciding who wrote something. The next deadline in the EU is 2 December 2026. It applies to systems that were already on the market. This may be a good moment to discuss how you tell employees and candidates that you use AI to create text.

OpenAI is adding invisible watermarks to text from its AI tools ChatGPT and Codex. This applies to users in the EU. The feature is called textGrain. The technology changes the choice of words slightly. A detector can recognise this pattern. A detector is a program that looks for a marker like this. The marker stays in the text if someone copies it and pastes it somewhere else. Readers cannot see the marker. It is built into a pattern in the text itself, not shown in a message alongside it. The rollout will take place over the coming weeks and applies to all subscriptions. The feature is available to users who are eligible. For HR, this matters for job adverts, internal messages and other text that employees create with AI. The marker makes AI-generated text easier to recognise. It does not prove it for certain. According to OpenAI, recognition falls from almost 100 per cent to less than 20 per cent if someone changes a quarter of the words. For now, the detector is available only to approved researchers and organisations. If the marker is missing, that does not prove a person wrote the text. So the marker alone is not a sound basis for deciding who wrote something. The next deadline in the EU is 2 December 2026. It applies to systems that were already on the market. This may be a good moment to discuss how you tell employees and candidates that you use AI to create text.