The web’s newest weapon against AI scrapers is a font
“ShieldFont” aims to poison AI training data without making pages unreadable for people.
In response to AI companies' habit of scraping vast portions of the public web to gather valuable training data, a duo of designers, Isaque Seneda and Gabriel Abrucio, have developed a new font aimed at thwarting these scraping practices. Titled ShieldFont, the creation was detailed in a white paper by the designers, which explains its purpose as offering web publishers a practical means to opt-out from unauthorized AI training and disrupt the data collection process when it goes unnoticed.
ShieldFont's innovation lies in its use of ligatures, a common feature in many fonts that traditionally replaces certain letter pairs with alternative readable versions when they appear adjacent to each other. However, ShieldFont takes a different approach by employing these ligatures to replace entire words with others, aiming to render the text devoid of any value to scrapers.
Crucially, this transformation occurs only when the font engine renders the page on the screen, ensuring that scrapers, which may simply download the plaintext source code, are presented with an altered version unseen by end users.
Written by urgent.news from Ars Technica's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.