New font aims to deter AI data scraping from the web

Designers Isaque Seneda and Gabriel Abrucio have developed a new font, ShieldFont, designed to prevent AI companies from scraping large amounts of public web data for training purposes. The practice has already led to lawsuits and technical fixes.

ShieldFont operates using ligatures, a font feature that typically replaces letter pairs with a more readable version. In ShieldFont, these ligatures are used to substitute entire words with others, thus destroying the text's value for scrapers. This substitution occurs only when the font engine renders the page, meaning scrapers downloading plain text source code receive an altered, nonsensical version that end-users never see.

The font is presented as a practical way for web publishers to opt-out of unauthorized AI training and disrupt what is collected when that choice is ignored. The alterations are invisible to human readers but render the text useless for machine reading.

This technical approach aims to provide web content creators with a new tool to defend against mass AI data collection without impacting the end-user reading experience.