2 days ago I was massively overcharged for the API service. I had a limit of USD $120 that was massively overrun to USD $340+.
Right away I tried to contact the Help Centre via the chatbot but I am not getting an answer.
Apparently the messages have not even been viewed. It’s been 48 hours.
Has anybody else had such an experience?
How long can I expect to wait for a response from OpenAI?
Welcome to the Forum!
There tends to be a small lag in how fast the back-end recognizes new charges despite a hard limit. This can unfortunately cause situation like yours where the limit is exceeded.
Have you already identified the root cause for the charges, i.e. was this due to your own model consumption or the result of a potential leak in your API key?
If the latter, then you should have replaced your existing API keys with fresh ones.
As for the customer service, it can take a few days or even a couple of weeks to get a response given the high volume of queries they receive.
Thanks for getting back to me Jen.
There was a limit of the usage of $120 and the cost within a few minutes rose to over $390. (I managed to disconnect my card at $340). If it was a few dollars I would understand but 3 times the amount limit is not reasonable, nor acceptable.
I had normal and a small usage of the API key. And I checked but it was not hacked and used by someone else.
Even it had been the $120 limit should have limited my liability.
Also my account was automatically and without my approval and no warning increased from Tier 1 to tier 2 to Tier 3 within a few minutes.
It is clear that safeguards that you provide to your clients for the usage of your tool are not working.
And I never received an answer from your Helpdesk even though it was urgent.
Therefore I ask you to reimburse the amount under the European consumers protection act.
Welcome @jdnmk8
Did you use the API key anywhere where it could have gotten out of your custody. e.g. 3rd party tool, client side code, sharing with someone else?
If so, I’d recommend revoking the API key(s).
@jr.2509 and I are community members just like you.
The best way is to just wait for support as they are the only ones who’ll be able to take any actions to help you.
Thank you Sukhman for your support.
Yes I deleted the keys and took off my card details from the system in a few minutes.
I was using the API to build a chatbot but I do not actually think the API was hacked. I think the OpenAI just multiplied queries within a few seconds/minutes on a massive scale within the app I was building…
My surprise and disappointment comes from the fact that OpenAI overran the usage limit massively and increase my user tier limit all this within a few minutes. That’s not normal. I signed up for a limited liability of USD $120.
If I had not seen the charges coming in by email it could have continued to 10000 dollars or more… All limits set by OpenAi were off!
The user tier increase is automatic and tied to expenditure. The purpose is to provide higher rate limits to those that spend a certain amount on the API, not to increase your self-imposed limits (the increase to your total monthly allowance is separate from the budget you set).
Can you see the usage that corresponds to these charges? That would be a good way to know if it was to do with your use or something else.
Additionally, I’m curious about the automatic charges… to my knowledge, they changed to a prepaid pricing model, so unless you have set up the automatic top-up, this shouldn’t occur regardless (it should stop when you run out of credits), and even then it should only top-up to a preset amount…
I had not prepaid. The amounts were debited automatically from my card. All the safeguards were bypassed. I am still waiting for a resolution from OpenAI 2 weeks later…
They don’t really read stuff.
Hopefully you will get a great resolution soon. I agree that seems unreasonable.
Was it o1 model? Otherwise, I think the rate limit should have also stopped you. o1 is challenging because we don’t know the cost before making the query, and it’s very expensive.
It is best to start building with the cheapest models and then upgrade as we confirm the actual costs. Sometimes the bills have surprised me as well. It’s easy to forget something in the calculations.
One potential fix could be for OpenAI to set the rate limits per user account so that it’s maximum is derived from user budget. For example, if I set budget $100 my rate limit would be set so that I can’t consume my monthly budget in less than 5 minutes. It would still be overly generous allowance to spend monthly budget in 5-10 minutes. Cool off period of a few hours between tier increases would probably be a good idea.
I agree it seems obvious to me. My limit was $120 and it jumped in a few minutes to $390…