Why does my opt-out not follow my data through a training pipeline?

Setting robots.txt or an ai.txt signal tells compliant crawlers to skip your content, but that signal has no mechanism to reach data already scraped and sitting inside Common Crawl snapshots or derivative training datasets. Once your writing or art has been aggregated, dozens of downstream fine-tuning pipelines can train on it without ever seeing your original refusal. The IETF AIPREF working group, chartered in January 2025, is building a machine-readable vocabulary for content usage preferences, but it applies only to future crawls and has no retroactive reach into existing corpora. EU AI Act GPAI obligations that took effect in August 2025 require providers to honor opt-outs, but no technical enforcement mechanism exists for multi-hop pipelines where data has already changed hands.