It hooks the send button (and Enter), checks the prompt on your device, and if it finds something, shows what and where. You can Send Anyway or Cancel; the paid tier adds Sanitize & Send (replaces values with placeholders) and scanning of attached files — PDFs, Word/Excel/PowerPoint, images via in-browser OCR, zip archives.
Things you'll probably ask:
- Where does my text go? Without an account, nowhere: the checks run in the extension, with the same rules our server uses (the extension's scanner is generated from the server's rule code). If you add an API key, text is sent to our API for scanning and your history keeps the first 500 characters with secrets masked. Image OCR also runs locally, in the extension (the OCR engine and language data ship inside it; nothing is fetched at runtime). - How does detection work? Pattern rules plus validators (e.g. Luhn for card numbers, context checks so "ISBN 978…" isn't a card). No LLM in the loop. It will miss secrets it has no pattern for, and it will sometimes flag harmless text — both are fixable and we'd like examples. - Why an extension and not a proxy? Consumer AI apps pin TLS and personal laptops don't have a corporate proxy. The extension sees the text before it's encrypted.
Free to install and free to use: prompt checks run on your device with no account and no limit. Personal ($5/month) adds one-click Sanitize, attachment/screenshot scanning, history and the Android app. Organizations can roll it out via Chrome policy (Google Admin/Intune). Would love feedback, especially false positives and things it should have caught.
https://eraseai.ai/start?utm_source=hackernews&utm_medium=post&utm_campaign=launch