Vet key-proxy services before trusting them
OpenAI key leaked three times via a key-proxy service · Security incident · X · @pawelkarniej · Sep 2026
An iOS developer says their OpenAI key leaked three times while using a third-party key proxy, and that usage limits kept the losses small. They no longer recommend that service.
Lesson
Treat any third party that holds your key as part of your attack surface: keep per-key limits, monitor usage and rotate on any anomaly.
Fix prompt
Copy this into your own AI tool (Cursor, Lovable, Claude Code, Bolt…). It's general guidance written by us. Review what the AI suggests before applying it.
I'm a beginner and want to make sure my API keys are safe in my OpenAI project. 1) Search the code for hardcoded keys or tokens and for secrets in variables that ship to the browser (NEXT_PUBLIC_, VITE_, REACT_APP_). List the file and line, but don't print the values. 2) For each one, tell me whether it's meant to be public (like a Supabase anon key) or secret. 3) For secret ones, show me how to move them to server-side env vars and call the provider through a server route. 4) Give me the steps to rotate each exposed key in the provider's dashboard and set a spending limit. 5) Make sure .env is in .gitignore and in my AI tool's ignore file. Don't change anything until I've approved your plan. Never ask me to paste API keys, passwords or .env contents; tell me where to check them myself.
Source: X ↗