PromptShield

Local sensitive-information detection for text prompts on ChatGPT, Gemini, personal Microsoft Copilot and Claude.

PromptShield — Privacy Protection for AI Conversations PromptShield helps you identify and sanitize sensitive information before sending text prompts to ChatGPT, Google Gemini, Microsoft Copilot, and Claude. Whether you're writing everyday messages, working with code, or using AI for professional tasks, PromptShield adds a simple layer of privacy protection. HOW IT WORKS When you send a text prompt, PromptShield checks it locally in your browser for supported sensitive-information patterns. If potential sensitive information is detected, you can choose to sanitize the message, send it unchanged, or cancel. Sanitization replaces detected values with placeholders before submission. KEY FEATURES Local Sensitive Information Detection Detection and sanitization are performed in your browser without sending prompt content to an external scanning service. Support for Four AI Platforms Protect text prompts on ChatGPT, Google Gemini, Microsoft Copilot (personal chat), and Claude. Custom Privacy Rules Create your own exact-text rules to identify information you want to protect. Rules can be edited, enabled, disabled, or deleted and are stored locally in your browser. Flexible Sending Controls Review detected information and decide whether to sanitize the prompt, send it unchanged, or cancel. Improved Phone Number Detection Recognizes additional supported phone-number formats, including common Australian mobile formats. Optional Local Response Restoration On ChatGPT, supported placeholders in AI responses can optionally be restored locally for easier reading. This feature is disabled by default and does not restore developer secrets. FREE AND PRO Free: 100 scans per UTC month. Pro: Unlimited scans for AUD 1.99 per month. PRIVACY AND LIMITATIONS PromptShield provides best-effort, pattern-based protection. It may not detect every sensitive value, and some text may be incorrectly identified. Detection and sanitization are performed locally. Subscription management uses ExtensionPay. If protection is unavailable, normal message submission remains possible. Users can also choose to send text without sanitization. PromptShield currently protects supported text prompts only. Attachments, images, and audio are not scanned. Take more control over what you share with AI using PromptShield.