Big AI labs back __usage_ai-training opt-out token

The announcement unites four labs that have often competed over how web content flows into AI systems. Behind the token sits an individual IETF Internet-Draft, "Machine-Readable Training Preference Signal", dated October 30, 2026, expiring May 3, 2027, with the AIPREF working group chaired by Cloudflare's Mark Nottingham and Cisco's Suresh Krishnan. The participant list — Meta, Cloudflare, the Common Crawl Foundation, News Corp, the Financial Times, the BBC, the Electronic Frontier Foundation and Creative Commons — shows the discussion extends well beyond the four labs.

Google's John Mueller framed the step as helping a site state AI training preferences across multiple providers at once, and Google plans to read an opt-out via the token as if the publisher had disallowed Google-Extended. Microsoft, for its part, intends to build on the shared preference while adding a separate setting for AI grounding, the retrieval step that pulls live content into chat answers rather than training weights.

Two limits remain. OpenAI describes the mechanism as a coordination tool rather than a security control under RFC 9309, so non-compliant scrapers can ignore it, and the opt-out does not pull material already gathered into datasets or baked into models. The timelines also differ sharply: Anthropic says it has begun implementation, OpenAI speaks of the coming months, and Microsoft and Google point to 2027.