Tenebrae Guardian

On-device AI safety: warns before you share sensitive data with a chatbot and flags scams in AI replies. Stays on your device.

Tenebrae Guardian is a seatbelt for talking to AI. It works on ChatGPT, Claude, and Gemini and everything it does runs on your own device. It's free, needs no account, and makes no network connections at all. Your conversations are never sent to us or anyone else. BEFORE YOU SEND Type freely. When your message contains sensitive data a Social Security number, a credit card number, a phone number, an email address, an API key, or a password or secret you've written out Guardian catches it and offers to hide it with one click, swapping the real value for a safe placeholder. Nothing is sent until you choose to send it. WHEN THE AI REPLIES Guardian scans the reply for scams: phishing links, requests for your passwords or codes, account scare-tactics, and other manipulation. If something looks off, you get a plain-English warning above the message. HIDDEN-INSTRUCTION PROTECTION Pasting a web page, email, or document into the chat for the AI to act on? Guardian checks it for hidden instructions secretly aimed at the AI concealed in comments, invisible characters, or encoding and warns you before you send. PRIVATE BY DESIGN All detection runs locally inside the extension. It makes no network requests of any kind, stores nothing about your messages, and has no account or sign-in. The only thing kept on your device is a small count of how many items it has hidden, shown in the popup. Free. No account. No ads. No tracking. We never see your data because it never leaves your device.