In-Browser PII Anonymization for AI Prompts

An email address can end up in an AI prompt simply because someone pasted a message or a document into the chat. I built a browser extension to mask those addresses before the request is sent.

The extension uses Manifest V3 and works in Chrome and Firefox. It intercepts ChatGPT requests in the browser and replaces detected email addresses with a placeholder. For example, an address in a pasted message becomes [EMAIL_ADDRESS] while the surrounding text remains available to the model.

My work spans the extension behavior and its interface, using TypeScript, JavaScript, and React. The important part of the workflow is where the masking happens: before the outgoing request, rather than after the text has already reached the AI service.

The project focuses on email redaction within this browser workflow. That gives the feature a concrete scope that can be explained and checked.

This is relevant to teams that need a browser extension, a change to an existing web workflow, or an additional check before information is passed to an AI tool.