Hidden text
Letters in the background colour, at a minimal size or off the page. You don't see them, but AI reads them.
Prompt injectionDetection in files
Is the file hiding malicious instructions for AI?
Upload the file and find out whether it contains hidden text, metadata or code crafted to deceive AI systems.
PromptSentry does not guarantee that every attack is detected. Each result states what was analyzed and what was left out.
What a person sees and what AI reads are not always the same document.
Metadata and comments
Internal file fields, old revisions and annotations that don't show up in print.
Disguised messages
Encoded instructions, such as Base64, or written with invisible and look-alike characters.
Received files are treated as hostile
Macros, scripts and JavaScript inside documents are never executed. Each file is opened in an isolated environment without network access, and the original is kept unchanged.
See security details ›