AvoPseudo

Pasting a client email, a contract extract, a patient list or an internal codename into ChatGPT is a two-second gesture with consequences you can't take back. AvoPseudo makes that gesture safe: it inspects the text at the exact moment you paste it, and replaces sensitive values with reversible placeholders before the text lands in the chat box. Not on send. Not after a warning you might click through. Before. Press Ctrl/⌘+V (or right-click → Paste, or a paste button) in ChatGPT, Claude, Gemini, Copilot and the rest. AvoPseudo reads the clipboard text, checks it against your rules, and acts: • Pseudonymise — each sensitive value is swapped for a reversible tag such as [EMAIL_a1b2c3] or [CLIENT_9f4e21]. What appears in the composer is already pseudonymised, so the model cannot see the raw data even if you send by accident. • Warn — the text is pasted unchanged, with a clear notice about what it contains. • Block — nothing is pasted at all. A change made silently inside a text box is a change you can't trust. Every time AvoPseudo steps in, a branded panel appears right above the composer and says, in words, that the message comes from your AvoPseudo extension — not from the website, and not from the AI. It lists every value that was replaced, masked, next to the placeholder that replaced it, with a "Show originals" toggle and an "Undo — paste my original" button if you disagree. All detection and pseudonymisation happen in your browser. Your prompts, the detected values, the activity log and the statistics never leave your device — there is no account, no server, and no telemetry. The only outbound request the extension can make is the one-time download of the on-device model, and only if you enable that feature. Contact : info@avotempo.be