Invisible text that AI chatbots understand and humans can't?

Researchers have revealed methods for inserting invisible text into prompts and outputs of AI chatbots like Claude and Copilot, exploiting Unicode quirks to create steganographic channels. This allows for covert communications and potential data leaks.

• Invisible Unicode characters can be used as a covert channel.
• Large language models can interpret these characters.
• Attackers can sneak in malicious instructions.
• Codes are invisible in browsers but recognized by AI.
• Mitigations were introduced after private reporting.