airescues / chatgpt / ChatGPT hub

ChatGPT by OpenAI

Leaked OpenAI API key: delete it, then cap what it can spend

A leaked OpenAI API key can lead to unexpected charges. Delete it, check what it was used for, and set a hard limit — spend alerts alone don’t stop the traffic.

When OpenAI spots it first

A key found on the public internet, or leaked inside an app in an app store, is disabled immediately.

So a key that suddenly stops working may already have been caught and disabled by OpenAI.

Shutting the key down yourself

  • Delete the affected key through the API key dashboard, review usage and contact Support.
  • After any account compromise, delete your API keys and review API usage.
  • Support opens from the chat bubble at the bottom-right of help.openai.com.

When the leak came with your OpenAI login

  • ChatGPT and the API Platform share one OpenAI account and one password, so a leaked password reaches both.
  • Change the password: all current sessions are logged out within 30 minutes.
  • In Active sessions choose Log out all, then Log out of all devices.
  • Active sessions does not list third-party app or Codex CLI sessions.
  • Turn on MFA only after those steps; it then covers the API Platform too.
  • A push approval for a login you didn’t start: tap «No, deny access».

Charges from requests you didn’t send

  • A card charge you don’t recognise: contact your bank immediately.
  • OpenAI must validate unauthorized charges before it issues a refund.
  • Report the activity on the Unauthorized Activity webform with «Report unauthorized activity involving my account».
  • Already disputing the charge with your card? No separate refund is issued while that dispute is active.
  • After a leak, be cautious with emails that ask for credentials or send you to pages that want account details.

Deleting the account will not reset the API

Deletion ends access to ChatGPT and the API alike.

It does not reset API phone verification: one number verifies the first API key for up to three accounts.

Alerts versus a hard limit

A hard spend limit stops API traffic; spend alerts alone don’t, and the total may slightly overshoot.

Two columns for a leaked OpenAI API key. With spend alerts only: traffic from the key keeps running and you get an alert. With a hard spend limit: traffic is stopped at the limit, and the total may slightly overshoot it.
Alerts tell you; only a hard spend limit stops API traffic — and it may slightly overshoot.

Where a key should never live

A key in a browser or mobile app can lead to unexpected charges or exposure of account data.

Sharing API keys is against the Terms of Use.

keys are not for support chats OpenAI says it does not request API keys, passwords or verification codes by email, text or chat — so paste a key nowhere except your own server. Check that any message «from OpenAI» comes from an official OpenAI domain.

Next for ChatGPT

Sources — OpenAI help pages, checked October 2026

AIRescues explains the official ways back into your own AI account. We are not ChatGPT, Claude, Gemini, Grok or DeepSeek and cannot open any account for you.

Facts are checked against each platform’s help pages; every guide shows the month it was checked.

How we check About Responsible use