AI giants agree on one signal: websites can now block their content from training AI

The four biggest AI companies — OpenAI, Google, Microsoft and Anthropic — announced on Thursday, October 30, 2026, that they will honour one shared signal allowing a website owner to block their content from being used to train AI models. It is the first time these large companies have agreed on a single mechanism of this kind.

The signal goes into a file called robots.txt — a small file that almost every website has, which tells data-collection programs (crawlers) what they are allowed to do. Until now, an owner who wanted to block AI training had to write a separate rule for each company, and sometimes for each data-collection program of a single company. The new signal changes that: one rule is read by all participating companies.

The signal comes from the work of the Internet Engineering Task Force (IETF) group working on AI preferences, and its technical draft was submitted on October 30. It is important to know that it is still a draft — not a formally adopted standard — but the four companies have already committed to using it.

There are three important limits to know. First, the signal covers training of generative AI models only, not other AI work. Second, it covers future training only: it will not remove content that has already been used to train existing models. Third, if an owner has not set the signal, their content remains allowed to be collected — blocking requires the owner's own action.

The timetable differs by company. Microsoft says it will start honouring the signal from February 1, 2027, and has also announced its own separate "AI Grounding" setting — which covers when an AI reads a page to answer a user's question, rather than to train a model. OpenAI says it will implement it "in the coming months" and that its existing robots.txt controls will continue to apply. Google is expected to follow sometime in 2027, and reports say Anthropic has already begun. Meta says it supports the idea, but has not committed to implementing it.

The announcement comes as publishers complain that AI programs collect their articles without permission or payment — and major cases are still before the courts, including the New York Times case against OpenAI and Microsoft filed in December 2023.

What does this mean for you?

Writers, journalists and publishers: This is a new tool to protect your work. If you run an online newspaper, a blog or a news site — even a site like this one, Pashavo — you can now set one rule instead of four or more. But remember: it does not help with articles already collected, and implementation takes months. If you want to block, do not wait until February.

Small businesses with websites: If you have a website for your shop, hotel or service, your product descriptions, photos and prices are content too. This signal gives you a voice: you can choose to let AI learn from your business — which may make it show up in AI answers — or to block it. Your choice, not the AI companies'.

Content creators and designers: If you write, photograph or make videos and put them online, your work now has a simpler shared protection. But this protection is voluntary for companies — it is not a law. A company that does not participate can keep collecting, so you still need to understand your country's copyright rules.

Everyday users: For the ordinary reader, this news means more transparency. Websites that choose to block training may appear less in AI answers, while those that allow it may appear more. When you read an AI answer, remember: it may have learned from content whose owner allowed it — or failed to block it in time.