OpenAI to honor a shared robots.txt signal for AI-training opt-outs
OpenAI says it will support a cross-provider robots.txt token, built on an IETF draft, letting sites set one AI-training preference for multiple crawlers.
What to know
- OpenAI will honor a new shared robots.txt token (__usage_ai-training) letting sites express one AI-training preference across multiple participating providers instead of separate per-crawler rules.
- The signal comes from an individual IETF Internet-Draft authored by Anthropic's Kevin Kelley; the IETF notes such drafts are not endorsed and have no formal standing in the standards process.
- The mechanism adds no new robots.txt syntax, keeps existing crawler-control directives unchanged, and allows private agreements to override the expressed preference.
- OpenAI has not yet implemented the signal, saying details will follow in the coming months, and is encouraging other AI providers to adopt it as well.
“User-Agent: __usage_ai-training”
draft-kelley-aipref-training-preference-signal-00, IETF Internet-Draft example robots.txt group · Unite.AI ↗ · Oct 29
OpenAI AI companyKevin Kelley Anthropic engineer, draft authorAnthropic AI company, IETF working group participantIETF AI Preferences working group Standards body
How it unfolded 2 developments, newest first · click a bar or a number to jump articlesposts
-
2
IETF draft by Anthropic's Kevin Kelley defines the new signal
Reporting detailed the underlying mechanism: an individual, not-yet-endorsed IETF Internet-Draft titled "Machine-Readable Training Preference Signal," authored by Anthropic's Kevin Kelley, defines a cross-vendor policy token (__usage_ai-training) that binds "no AI training" preferences to paths using existing robots.txt syntax, with no new grammar and existing parsers unaffected.
“no AI training…”
— Machine-Readable Training Preference Signal draft -
first by Unite.AI, 12d ago
-
-
1
OpenAI commits to honoring a shared AI-training preference signal
OpenAI announced it will honor a shared robots.txt signal so site operators can communicate one AI-training preference to multiple participating providers instead of per-crawler rules, calling the step voluntary and saying it will implement the signal in the coming months.
Also covered reported alongside — the timeline has no entry for these yet
-
first by OpenAI, 12d ago · also OpenAI News