same here, vector database was fine for a while, all the health checks are good - but it hangs when contacting render…tried everything…glad I found this so I can stop wasting time.
Hi everyone,
Thank you for working through this with us - we're urgently investigating this now.
For an immediate unblock, please set the GPTs "Privacy settings" to "Always allow" for any impacted GPTs you are attempting to use.
For convenience, here's a link to the status page for this issue: https://status.openai.com/incidents/01K2Z3WTYEM0WN3EPJMY25VD13
We'll keep the status page and this thread updated!
From my experience this doesnt fix the issue. I have this already set on my GPTs to no avail. Please let us know as you have updates. thx
This response is not valid. Gets have been working with always allow. The issue stems from any type of post/patch action which even with “always allow” enabled required explicit authorization and approval each time. This does not solve the issue.
Since the issue arose the problem is that we do not receive a response to approve an action related to a patch/post. Changing the privacy settings does nothing for that.
^ shared with the team, thanks @Bryan_Weinstein and @michael.voll! Will keep this thread updated as I learn more.
Your ‘descriptiob’ of the issue is also incorrect. It’s not after ‘permission is given’ that it’s stuck, it gets stuck before then.
Thank you for the note. We've updated the initial description!
I’ve just started seeing traffic coming through to my API so looks like they’ve fixed something, but now it’s stopped again
It seems to be working again for now. Thank you!
It seems to be working properly FINALLY!!! THANK YOU!!! This took far, far, far too long to recognize and correct…. In the future how can we notify the OpenAI staff of “critical” issues such as this?! The standard way of contact is unacceptable for critical issues…
Finally working on my end as well! Thanks team 
Glad to hear it's working for you all now and apologies for the time it took for us to jump in.
Going forward, we'll work on ways for community forum threads to trigger alerts on our end and are always adding more internal alerts in general.
Please do continue adding comments in these types of threads and writing into support@openai.com.
Thanks again for your patience here!
Finally, it’s working on my side too! Huge thanks to the team for fixing this issue — the past three days have been a nightmare.
I tested and GPT Actions with the authentication type ‘OAuth’ are now working when either ‘Ask’ or ‘Always allow’ is selected in GPT’s Privacy settings.
GPT Actions with the authentication type ‘API Key’ (when ‘Auth type’ is set to ‘Bearer’) also work when either ‘Ask’ or ‘Always allow’ is selected in GPT’s Privacy settings.
It looks like the mitigation measures have fixed the bug!
Huge thanks!!
It’s working again for me now. Many thanks to the development team for quickly fixing the problem.
Working for me now, thaks Dev Team!!
MAJOR BUG STILL REMAINING
On the exact day the authorization bug occurred (16th / 17th of august) another bug hit the surface.
Can be a coincidence, but it has also to do with the actions, within a custom GPT.
When your GTP receives any data from a custom action, it has NO ACCESS to it, when it does not explicit use it the very 1st message after it receives it.
It starts hallucinating when you try to access the retrieved data.
I have debugged this for about 40 hours, 3 days, both GPT 4 and 5.
TEST SETUP
-
GPT receives data (see below) and is instructed to output only the 1st sentence
-
data :Foo bar!
How are you?
Hello world -
It outputs (as instructed): ## Foo bar!
-
When you ask “what’s the second line, or the third line, or all lines” it starts hallucinating.
And when it only says something like “received the data - what’s next?” it has no access to any of the received data, because it did not output it in the very 1st message after it received it.
Custom GTP can not access (remember?) any received data, fetched from an action, when it is not using (outputting) it the very 1st message after it did receive it.
SEE SCREENSHOTS / INFO