News event: October 30, 2026. Website publishers could soon have a simpler way to communicate whether their material may be used to train generative AI. OpenAI, Google, Microsoft and Anthropic have signaled support for a common preference mechanism that can be expressed in a site’s existing robots.txt file.
What the proposed AI training signal does
The proposal, published on October 30 as an Internet-Draft by Anthropic standards contributor Kevin Kelley, introduces a shared policy token for expressing preferences about AI training. A publisher would no longer need to rely entirely on separate company-specific controls to communicate the same preference to multiple participating AI developers.
Crucially, the draft is a work in progress, not a finalized internet standard. It describes a content-use preference, rather than a technical lock that stops a crawler from fetching a page. Website owners can still use ordinary crawler-specific robots.txt rules to manage access, while the new token would address a different question: what a participating service should do with material it has obtained.
Which companies plan to honor it?
According to reporting published by The Verge on October 30, Anthropic has already adopted the signal. Google and Microsoft plan support during 2027, while OpenAI expects to implement it in the coming months. Meta has expressed support for common controls but has not committed publicly to honoring this particular signal.
Microsoft also intends to distinguish between AI model training and AI grounding, which uses sources to support generated answers. That separation matters to news websites: publishers may want their pages to remain discoverable in ordinary search while limiting certain downstream uses.
Why this matters for independent publishers
Smaller websites often lack the legal or engineering resources to maintain separate instructions for every AI crawler. A cross-vendor preference could reduce administrative work and clarify expectations. But adoption will depend on actual implementation, and publishers should not assume that adding the proposed token guarantees compliance by every company or prevents unauthorized scraping.
The next milestones are standards discussion and the participating companies’ rollouts. Until then, site operators should review existing robots.txt settings, search visibility requirements and any licensing arrangements before making changes.
Sources: The Verge, October 30, 2026; IETF Internet-Draft, October 30, 2026.
Context matters. Read about our editorial standards and corrections policy.
