How to stop screenshots with personal info going to AI tools

Why screenshots are the blind spot

Most checks look for patterns in text. A screenshot is pixels, so a card number in an image slips straight past a text filter. People paste screenshots into ChatGPT, Claude and Cursor all the time because it is quicker than typing out an error.

What ends up in them

• Email addresses and customer names in a mail client or CRM.
• Card numbers and billing details on an account page.
• API keys in a terminal window or an .env file.
• SSNs and personal details in a form or a PDF.
• Filenames that give the game away, such as jane-doe-ssn-tax-return.pdf.

What to look for in a control

• It reads images and PDFs, not only typed text.
• It runs on the device, so the screenshot is not sent to a third party to be checked.
• It checks filenames as well as contents.
• It uses the same block or redact rules you set for text.
• It tells the person what happened instead of failing silently.

Where Inveil fits

Inveil reads images, screenshots and PDFs with built-in OCR on the machine. No AI model and no outside API is involved, so nothing leaves the device to be checked. Filenames get the same check, and the rules are the ones you set for prompts: stop the request, or strip the details and let it go through.

Inveil is data loss prevention (DLP) for AI. It checks prompts, files and screenshots on each employee's machine before they reach an AI tool, blocks or redacts sensitive data by your rules, and logs every event.