tech
The web’s newest weapon against AI scrapers is a font
“ShieldFont” aims to poison AI training data without making pages unreadable for people.
TL;DR
- ShieldFont is a font designed to prevent AI companies from scraping public web data for training.
- It works by using ligatures to replace words in the HTML with unrelated words, which are only visible to scrapers.
- This alteration makes the text nonsensical for AI training but remains readable for human users.
- The font replaces approximately 24.5% of all words on a page and 45.8% of content words.
- Testing showed over 90% of pages were rejected by scraper quality filters after using ShieldFont.
- Pages that are still accepted contain "training-time garbage" that asserts nothing true.
- Potential side effects include issues with search engines, screen readers, and copy/paste tools.
- ShieldFont is not foolproof, as AI could use optical character recognition on rendered images.
- The creators aim to make large-scale scraping more expensive and less useful, enforcing the principle of consent for AI training.
- The goal is to encourage diverse methods for showing different content to humans versus machines.