Audit of 14,000 domains shows the AI data commons in rapid decline

BlancheMinerva · x · 2026-10-17

Stella Biderman points to "Consent in Crisis," the first large-scale longitudinal audit of consent protocols across 14,000 web domains underpinning AI training corpora. AI-specific restriction clauses have proliferated, websites' robots.txt and ToS signals are inconsistent, and previously open sites are closing to AI crawlers. The charts show restrictions on crawlers like GPTBot and Google-Extended surging through 2023–2024; by April 2024 OpenAI's crawler was restricted by 25.9% of domains and Anthropic's by 13.3%.

Original post →

More from Safety

Safety channel →