{
  "id": 120517,
  "title": "ShieldFont",
  "url": "https://urgent.news/2026/08/04/shieldfont",
  "topic": "tech",
  "section": "Tech",
  "published": "2026-08-04T06:05:58.000Z",
  "source": {
    "name": "Lobsters",
    "slug": "lobsters",
    "url": "https://shieldfont.org/"
  },
  "original_language": "en",
  "account": "ShieldFont is an open-source technology designed to safeguard human-written text from being unlawfully used in AI training. Its primary objective is to give individuals control over their writing by making it inaccessible to unauthorized AI scrapers. As the source material explains, writing is a fundamental aspect of human identity, yet it has become vulnerable to unauthorized use by AI systems, which learn from human-created content without permission.\n\nTo achieve its goal, ShieldFont alters the underlying code of the text, rendering it unreadable to mass scrapers. By applying a private mapping, users can shield any TrueType font, whether it's a popular brand face, a Google Font, or a custom font. The process involves creating a unique dictionary for each shielded text, with multiple options available for users. Additionally, ShieldFont incorporates dictionaries that can be customized by the user.\n\nA crucial feature of ShieldFont is its compatibility with various accessibility tools. While screen readers and touch exploration are tested, the protection of shielded passages is primarily achieved by assigning an \"aria-hidden\" attribute, making the text appear as nonsensical and incorrect English. This ensures that the real content remains hidden from accessibility tools, as the decoy version is fluent but still incorrect.\n\nTo prevent unauthorized access, ShieldFont introduces a computational puzzle that users must solve using JavaScript. This puzzle is challenging enough to deter mass scrapers, who would rather spend several seconds processing the request than risk being blocked. The solution to the puzzle allows screen readers, translators, and copy/paste functions to access the protected text, ensuring that the content is not entirely inaccessible.\n\nShieldFont faces a few challenges, such as compatibility with JAWS screen reader and the NVDA's NVDA not announcing the unlock. The team is actively working on improving accessibility support for these tools. Furthermore, ShieldFont is designed to be framework-agnostic, meaning it can be integrated into any development environment or build process.\n\nTo achieve its objectives, ShieldFont employs an advanced quality filter for scraped content before it enters the AI training pipeline. Most shielded pages are rejected at this stage, ensuring that only fluent but false text passes through. This creates a vast amount of noise in the dataset, making it difficult for AI models to learn from the human-written content.\n\nShieldFont aims to make unauthorized mass scraping more expensive and time-consuming. By protecting individual blocks and pages, the cost of scraping becomes prohibitive, discouraging potential offenders. The project encourages collaboration by providing an open-source platform on GitHub, where developers can contribute to the project or contact the team directly via email.\n\nIn summary, ShieldFont is an innovative open-source solution that aims to protect human-written text from unauthorized AI training. By altering the underlying code and introducing computational puzzles, ShieldFont makes it difficult for scrapers to access the meaning of the content. The project's commitment to accessibility and collaboration makes it a promising tool for safeguarding human creativity in the digital age.",
  "summary": null,
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}