# Labs Converge on Shared AI-Training Opt-Out Standards

Frontier labs will keep agreeing on shared, machine-readable consent signals for AI training ahead of regulation. Consent infrastructure (crawler tokens, licensing flags, provenance markers) becomes a recurring negotiation front between labs and publishers, and the pool of freely crawlable training data shrinks as opt-outs become easy to honor and enforce.

- Conviction: 38 / 100 (forming)
- Horizon: Emerging (watchlist)
- Tracking since: 2026-10-31T00:00:00.000Z
- Last updated: 2026-10-31T06:29:50.047Z
- Canonical: https://polylog.news/ai/trends/ai-training-consent-standards
- Publisher: Polylog

## Recent evidence

- [confirms] OpenAI, Google, Microsoft and Anthropic Agree to Honor One Robots.txt Token for Opting Out of AI Training (2026-10-31): OpenAI, Google, Microsoft and Anthropic agreed to honor one robots.txt token for opting out of AI training. Anthropic says it has begun implementing the signal, Microsoft targets February 1, 2027, and Google says only 2027.