Unexplained spending on API and other models

API Bugs
,
1

I’ve been running some applications with the ChatGPT API for some time. There was never an excessive expense and we always controlled outputs and costs.

However, recently, expenses have increased considerably, including models that we have never used in our application and a number of tokens that are completely outside the standard of what we use.

  • We revised our entire application and we do not use any models such as: gpt-4 or gpt-3.5-turbo.
  • We created a new api key but the excessive spending continued

I have no idea how to solve this problem. I contacted support but no response so far.

2

Does that include never using OpenAI API credentials within client code, never making OpenAI calls direct from application, but instead using only your own API proxy server to manage application user accounts and make calls to OpenAI?

That is the typical way for API keys to continuously be exposed and exploited when revoked and reissued. You cannot keep a secret by distributing it to user devices.