• Designers from S&A and type foundry Playtype have released a new typeface called ShieldFont, which poisons text read by web scrapers.
• Unlike visual tricks against AI scrapers, ShieldFont works at the source-code level, delivering normal text to human readers and contextual nonsense to robots.

Publishing your work, even for free, doesn’t grant anyone a license to copy it and profit from it. But that doesn’t stop companies like OpenAI, Anthropic, and X (formerly known as Twitter) from scraping the web to use it in products like ChatGPT, Claude, and Grok without authors’ permission.

On the visual front, efforts have been made to poison these AI databases using tools like Nightshade. Still, writers also have a need to protect themselves from the tech bros’ “move fast and break things” attitude and disregard for things like consent, ethics, and morals. The latest typeface from designers Seneda & Abrucio and type foundry Playtype, called ShieldFont, aims to proactively enforce consent by poisoning AI training.