AI Prompt Leak Guard

Checks AI prompts and files on your device for secrets, personal data and ID numbers before you send. Prompts stay on your device.

Check for sensitive text before you send a prompt to an AI chat site. AI Prompt Leak Guard checks prompts and files on your device, in your browser. Before a prompt is sent, it shows what it found, why it matters and what you can do. Prompt text is not sent to CandidLoom. WHAT IT DETECTS - API keys and access tokens from common AI, cloud, payment and developer services, plus JWTs, private keys, configuration secrets and database URLs with passwords. - Credentials in configuration files, command lines, request headers and signed links, plus crypto wallet seed phrases and keys. - Personal data: email addresses, phone numbers, labelled names and addresses, and tables of personal data. - Indian identifiers such as Aadhaar, PAN, voter ID, driving licence, GSTIN, UPI ID and IFSC, plus US, UK, EU and other national IDs, IBAN and payment cards. - Health and other sensitive details, insider-information markers, confidential business language and your own custom terms. WHAT YOU CAN DO - See what was found and why it matters before a prompt is sent. - Replace detected values with placeholders, preview the result first, or undo it. - Send once anyway where your settings allow, with an optional reason. - Pause protection on a site for a short time. It resumes on its own. - Redact selected text and copy it to use anywhere. - See from the toolbar whether prompts on the current tab are being checked. - Review counts of what protection did over the last day, weeks or months. PRIVACY All checks run on your device. Prompt text, detected values, file contents and page addresses are not sent to CandidLoom. Settings and activity counts stay in local extension storage. SUPPORTED SITES Works on popular AI chat websites. See the list at https://candidloom-hub.github.io/ai-prompt-leak-guard/supported-sites.html PLUS (OPTIONAL) Plus checks the text inside files, PDFs and images on your device before upload, and adds policy presets. One Plus key works in up to 3 of your browsers. Plans and prices are on the website. Detection can miss sensitive values or flag harmless text. Review your prompt and follow your organization's rules before sending.