This has happened 3 times and the result is a full block on API access until it magically and mysteriously resolves after about 2-4 days. The typical cycle is support takes about the same amount of time to respond, then closes the ticket since it’s not actively happening without investigation or providing any feedback on why it happens or how to resolve.
Is there some way others have found to avoid the downtime from the following?
Here’s the pattern:
The billing system locks me out. Does not auto-renew and any time I try to pay with an existing credit card or a new one I get a declined error.
The card account shows zero transactions that were declined. Implying that OpenAI is never attempting to bill the card. The cards are good with ample available credit and actively working for other vendors.
It is not browser cache or vpns or bad information for the card(s).
No action on my part has any effect for positive or negative. I just wait. Then it resolves and starts billing again.
I am having the exact same issue- I had a declined payment last night while running some API calls, and now it will not allow me to make a payment or update new payment info. I reached out to support hours ago, but no one has as much as read my messages.
I read the help page and followed the instructions, but they do not help. I have tried everything I have seen online that works for others (including trying to pay through mobile), to no avail.
We use OpenAI’s API in our commercial scale NLP pipeline on quick-turn projects, so for us to lose access for several days at a time, with a totally unresponsive support team, is a totally unacceptable business practice.
Update: As of this morning, I was able to process payment successfully. I did not hear back from support (messages are still labeled as unread), so possibly the issue fixed itself or it was some kind of 24 hour lockout.
I wanted to update with some new information. Support got back to me after a week or so. The issue had auto-resolved by then…however
The support tech noted that their payment system had disabled charging due to “more charge attempts than allowed on February 28-29”. This disables charge attempts “for 2-3 days”.
I’ve had auto-recharge enabled. I suspect there’s a bug in their auto recharge billing. So when you go negative, they keep retrying and exceed the number of attempts. Leading to the declined issue we see for 2 or 3 days until they reactivate.
If this is the case, disabling auto-recharge and precaching a sufficient amount for the expected load might avoid the downtime. Worst case, if we forget to replenish the card we’d get failed API calls but can quickly reallocate funds rather than suffering a 3 day blackout.
I’m getting ready to hit the api pretty hard which tends to trigger the recharge. I’ll report back if it avoids the issue.
It looks like OpenAI has a guard in place where you can only charge a billing source so many times within a single time period. If you exceed this rate, they disable your billing for a cool down period of about 2 to 3 days.
In my case, I had auto-billing enabled. During periods of high API load, the auto-billing was low enough and the charge rate high enough to exceed this threshold and trigger the cool down period.
However, there’s no visibility into what’s happening for the customer other than a rather misleading error of “transaction declined”.
You can avoid this in two ways:
Set a very high auto-billing amount to reduce the auto-billing frequency during high load
Disable auto-billing and manually refresh.
I think the benefits, drawbacks of both are pretty obvious. But either one will prevent the lockout from occurring.
“transaction declined” may be the best result you could get. In my case, my account get banned next day after my payment. I lost my money and my account. And the support team just don’t reply for a month, then they said they will investigate in the future, then again, no response for a week.
OpenAI is a LIE ← That’s true.