Siphrix — AI Action Monitor

Records what AI web apps do with your data and warns you about the risky parts. Audit-first: nothing is ever blocked.

Siphrix puts a policy check between you and AI web apps. When you upload a file or paste something that looks like a secret (API key, token, PEM block) into ChatGPT, Claude, Gemini, Copilot or Perplexity, Siphrix asks your LOCAL Siphrix engine for a verdict and — if your policy says no — holds the action back and shows a banner. • Local-only: the extension talks only to the Siphrix service running on your own machine (127.0.0.1). Nothing is sent to remote servers. • Privacy-first: it never sends file contents or pasted text anywhere. Secret detection happens locally; only a small piece of structured metadata (the site, an action type) reaches your local service. • Consent-first: protection is OFF until you turn it on, with a clear on/paused state. • Your rules: sign in to your Siphrix account to choose, per AI app, how strict to be (Off / Basic / Strict) or add your own rules.