PrAIvacy: Redact Sensitive Data in AI Prompts

Detect and redact sensitive data in your ChatGPT prompts — and 10 other AI sites. Runs on your device, not on a server.

Stop pasting secrets into ChatGPT. PrAIvacy spots the sensitive data in your prompts — names, emails, phone numbers, addresses, account IDs, and secret keys — and lets you redact it in one click before you hit send. Built for ChatGPT. It works the same way on 10 other AI sites. See the full list at praivacy.info Private by design: • On-device — detection runs on your device with a local AI model. PrAIvacy never sends your text to a server. • No account, no tracking, no analytics. • One-click redact — sensitive spans are replaced with safe placeholders before your message is sent; copy the AI's reply and your originals come back. • Flag your own terms — select a client name, project codename, or internal system in the prompt box and it is flagged everywhere from then on. Your list never leaves your device. • Minimal permissions — access only to the AI sites it supports, never to all websites. Using AI at work? Privacy laws such as GDPR and CCPA, and most company AI policies, ask you not to share personal data with services that don't need it. PrAIvacy helps you catch it before it leaves, and collects no personal data itself. It is a tool, not legal advice: detection can miss things, so keep using your judgment. PrAIvacy helps you protect your own and others' privacy when using AI tools. It does not bypass, disable, or circumvent any safety features or usage limits of those AI services. PrAIvacy is an independent project and is not affiliated with or endorsed by OpenAI or any other AI service it works with. Setup downloads the on-device detection model once from Hugging Face (about 810 MB, or 920 MB on computers without GPU acceleration); it then works offline, and model updates download in the background. Privacy policy: https://praivacy.info/privacy