Life.ru says ChatGPT's hidden text marks are for EU users under the AI Act; Vedomosti's headline names no limit and no law

What is reported. Life.ru says OpenAI "will start adding invisible watermarks" to ChatGPT's text answers "for users in the European Union". It gives the reason as meeting the requirements of the EU AI Act. It names no OpenAI statement and no date.

Where the accounts differ. Vedomosti's headline is broader: "ChatGPT will start marking generated texts with invisible signs." We hold only the headline, not the body, so we cannot tell whether Vedomosti's report limits the change to the EU.

	Life.ru	Vedomosti
Who gets the marks	EU users	Not stated in the headline
Reason given	The EU AI Act	Not stated in the headline
Source named	None	Body text not held
Start date	None	Not stated in the headline

Whose voice is missing. Neither report quotes OpenAI directly, an EU regulator, a ChatGPT user or an outside expert. Life.ru states the motive as fact rather than attributing it.

What is NOT settled. - Whether OpenAI has announced this itself. Neither report cites a statement. - Whether the marks apply only in the EU or everywhere. - When the change starts. - How the marks work, and whether they survive editing or translation. Neither report says. - Whether both reports rest on the same single source.

What this is about. Whether the EU's artificial-intelligence law is making OpenAI change ChatGPT, the most widely known AI chatbot, for European users.

How it got here. OpenAI, a US company, launched ChatGPT in November 2022. Within months, schools, publishers and governments were asking how anyone could tell machine-written text from human writing. The European Union was already drafting a law on AI. The AI Act was adopted in 2024 and entered into force on 1 August 2024. Its duties take effect in stages over several years rather than all at once.

One strand of the law deals with transparency. Providers of systems that generate text, images, audio or video must make the output detectable as machine-made, in a form a machine can read. This is what "watermarking" refers to: a hidden signal in the output that software can detect even though a reader cannot see it. Marking text is generally seen as harder than marking images, because text is easy to reword.

The actors. OpenAI builds and runs ChatGPT. It must follow EU law to offer the service to users in the EU's 27 member states. The European Commission and national authorities enforce the AI Act. The two outlets reporting this are Russian. ChatGPT is not officially offered in Russia.

What is at stake. If a dominant AI company changes how its product works because of an EU law, that shows Brussels can set rules for a US platform. If the change is EU-only, users elsewhere get a different product. If it is global, one region's law ends up setting the standard for everyone. Whether text watermarks can be detected reliably also matters to teachers, publishers and anyone judging whether a piece of writing is genuine.

Where it stands. On 5 October two outlets reported the plan. Neither cites an OpenAI announcement or gives a date. The EU scope rests on one report.

What is being argued about

Whether OpenAI's invisible text marks apply only to EU users or to every user
Whether the EU AI Act is the stated reason, or an outlet's reading of it
No OpenAI statement, start date or method is cited in either report
No regulator, user or outside expert is quoted

Two reports, two scopesSide by side

Outlet	What it said	EU-only?	Law named?	Source named?
Life.ru	OpenAI will start adding invisible watermarks to ChatGPT's text answers for users in the European Union, to meet the EU AI Act	Yes	Yes, the EU AI Act	No
Vedomosti	ChatGPT will start marking generated texts with invisible signs (headline only; body text not held)	Not in the headline	Not in the headline	Body text not held

Quotes translated from Russian. We hold Vedomosti's headline but not its body text.

Sources L Life.ru V Vedomosti

The bias in this coverageBias

OpenAI's account, unchecked

The one report with body text gives OpenAI's plan and its motive as fact, and quotes no regulator, user or outside expert. The other carries a headline only, and the headline leaves out that the change is for EU users. Neither gives a start date.

0 reports quoting a regulator, user or expert 1 headlines saying the change is EU-only 1 reports with body text 0 reports giving a start date

Scored without the mastheads. Where the two passes differ, both are kept.

Sources L Life.ru V Vedomosti

Claims and verdicts

No claims registered.

Who wrote it

А Анна Суслова Vedomosti

What the bias passes found in their reporting

Attribution inside the headline the same claim carried as the speaker's, not as fact×1

Framing Frames the finding as the EU contradicting Kyiv. The report has no body text to support that.×1

Headline carries a claim as attribution "strikes on civilian sites" is stated in the headline, the colon doing the work of attribution×1

Headline omission The headline drops both the EU scope and the law, so the change reads as applying everywhere.×1

Claims we checked that rest on their reporting

3 unverified2 true

From 569 reports we hold under this byline, at Vedomosti. Two independent readings, each made without the masthead, a quote attached to every tag. A record, never a rating. Their full record →

Л Лия Мурадьян Life.ru

What the bias passes found in their reporting

Actor framing The company is the only actor named. The regulator appears only as a law it obeys.×1

Asserted versus attributed the first sentence states the talks went 'excellently' as fact; the attribution to Trump comes in the next sentence×1

Company claim stated flat OpenAI's plan is stated as fact, with no statement cited and no date×1

Hedged versus stated states flat that the talks went 'excellently' and only attributes it to Trump in the next sentence×1

From 724 reports we hold under this byline, at Life.ru. Two independent readings, each made without the masthead, a quote attached to every tag. A record, never a rating. Their full record →

The bylines on the reports this story is built from. Hover one for what our passes have measured on their reporting.

Coverage, outlet by outlet

V Vedomosti —

ChatGPT начнет помечать сгенерированные тексты невидимыми знаками

Its headline says ChatGPT will mark generated texts with invisible signs and names neither the EU nor the law. We could not read its body.

ChatGPT в ближайшие две недели начнет помечать сгенерированные тексты невидимыми водяными знаками, следует из сообщения OpenAI. «В ближайшие недели мы добавим невидимые водяные знаки в соответствующие текстовые ответы ChatGPT и Codex в Европейском союзе», – говорится в сообщении. С 5 октября все пользователи нейросетей смогут включить отображение текстовых водяных знаков для некоторых моделей ИИ (данная опция по умолчанию отключена). Также компания предоставит доступ к детектору текстовых водяных знаков. В индивидуальном порядке доступ получат исследователи и экспертные организации. При этом о

Headline scope: the headline drops the EU-only limit, so it reads as applying to every user — “ChatGPT начнет помечать сгенерированные тексты невидимыми знаками”

Headline omission: The headline drops both the EU scope and the law, so the change reads as applying everywhere. — “ChatGPT начнет помечать сгенерированные тексты невидимыми знаками”

Read it at Vedomosti → https://www.vedomosti.ru/technology/news/2026/10/05/1234492-chatgpt-nac…

L Life.ru —

OpenAI добавит невидимые метки в тексты ChatGPT для пользователей из ЕС

The only report we hold with body text. It limits the marks to users in the EU and gives compliance with the EU AI Act as the reason, without saying who gave that reason.

Новая система получила название textGrain. Она не добавляет в текст скрытые символы или специальные пробелы. Вместо этого модель немного меняет статистику при выборе слов и их частей. Благодаря этому в тексте формируется незаметный для человека паттерн, который способен распознать специальный детектор. В ближайшие недели маркировка появится в подходящих текстовых ответах ChatGPT и Codex для пользователей из ЕС. OpenAI включит функцию на всех тарифах. Запускать функцию за пределами Евросоюза компания пока не планирует. OpenAI пока не станет открывать публичный сервис для проверки таких текстов.

Company claim stated flat: OpenAI's plan is stated as fact, with no statement cited and no date — “OpenAI начнёт добавлять невидимые водяные знаки”

Motive stated flat: compliance with the AI Act is given as the reason, without attribution — “чтобы выполнить требования европейского закона об искусственном интеллекте EU AI Act”

Actor framing: The company is the only actor named. The regulator appears only as a law it obeys. — “OpenAI начнёт добавлять невидимые водяные знаки”

Motive stated flat: Compliance with the EU AI Act is given as the motive, without being attributed to anyone. — “чтобы выполнить требования европейского закона об искусственном интеллекте EU AI Act”

Read it at Life.ru → https://life.ru/p/openai-dobavit-nevidimye-metki-v-teksty-chatgpt-dlya-…

Read another desk

Sections

Registers

Desks