"The requested action requires approval"

GPT builders Plugins / Actions builders
1

Hi guys

I’m encountering an issue when testing a custom GPT with actions calling our public API
Here’s a breakdown of the situation:

Request sent by GPT:

json
Copy
Edit
{
“SessionId”: “777af0a0-bc17-418e-8480-c20fa5e00c00”,
“Emails”: [“aXXXX@pXXXX.com”],
“Cellphones”: null,
“GroupIds”: null,
“IncludeExtraFields”: false,
“IncludeGroups”: false,
“IncludeCountryState”: false,
“SmsStatus”: null,
“EmailStatus”: null,
“PageNumber”: null,
“PageSize”: null
}
Response from our API:

json
Copy
Edit
{
“SessionId”: “777af0a0-bc17-418e-8480-c20fa5e00c00”,
“Status”: “Success”,
“Error”: null,
“Count”: 1,
“TotalCount”: 1,
“PageNumber”: 0,
“PageSize”: 10000,
“Clients”: [
{
“cellphone”: “05XXXXXXXXX”,
“cellphoneStatus”: 0,
“email”: “aXXX@pXXXXX.com”,
“emailStatus”: 1,
“firstName”: “XXX”,
“lastName”: “XXXXXX”,
“telephone”: “05XXXXXXXX”,
“address”: “דשגדש”,
“company”: “pulseem”,
“city”: “דשגדש”,
“extraData”: null,
“groups”: null,
“extraFields”: null
}
]
}
But instead, the GPT receives this response:

json
Copy
Edit
{
“function_name”: “GetClientsData”,
“domain”: “XXXXX”,
“message”: “The requested action requires approval”,
“action_id”: “g-772266a65ba6c07970eb36b405afafb43ebf35a7”
}
We want to clarify:

We do not block or restrict access to this endpoint — the API is publicly available.

The response seems to be generated by OpenAI, not our API.

Could you confirm what kind of approval this refers to? Is this a domain or endpoint review that OpenAI enforces?

How can we request approval or whitelisting of this action so that our GPT can function normally?

I have tried everything written here, privacy on “always”, publishing the GPT
but the problem wasn’t solved

1 Like
2

Hi, @Pulseem. Welcome to the forum.

It seems like that there something has changed recently. For a little while it was cut off. And you couldn’t do something until you had a higher tier account.

BUT, now you have to set up your builder profile. In the settings.
If you got any difficulties ChatGPT itself (4o) should also be able to help you solve that for you. At least when using websearch.

A settings window displays a prompt to verify a new domain by entering an address, with options to submit or cancel. (Captioned by AI)
A settings window displays a prompt to verify a new domain by entering an address, with options to submit or cancel. (Captioned by AI)1743×1514 165 KB

3

Hi, @hugebelts
Thank you!
I actually already verified my domain before testing it, but it still doesn’t work
You got any other ideas other than going to the upper tiers?

4

The thing is depending on the domain you’re using and how you’re doing it there may be other entries. Maybe a DNS entry for your domain or whatever. I can’t specify more precisely. Because there’s soooo many possibilities. You know?

It could also additionally how the json or yaml or what you’re using is not correct.
OR, it needs an authentication method or another. Just try. Ask ChatGPT.

Or even another AI. I do this myself the same way. Even though I got a 20 year programming background. I could figure it out manually as well. Yes. Give your team mate ChatGPT a chance. LOL.

Good Luck. If it’s possible to solve you’ll figure it out.

P.S.: One additional tip:

Try another API first, maybe even one of the examples of OpenAI, if you get the same error, solve that first.

5

hey @Pulseem @hugebelts

I’ve been trying to figure this out too for 24 hours lol!

I’d like to share what helped me:

  1. Go to your Custom GPT Home (not in the configuration)
  2. Click the the name of your custom gpt and you’ll find Privacy Settings then click it
  3. A modal will pop and you’ll see you’re URL where you can toggle between (“Ask” or “Always Allow”) → by default it’s “Ask” that’s why its always asking “the requested action requires approval” after setting it to “Always allow” I can now talk to the API freely!

hope this helps!
Cheers

2 Likes
6

So I was having the same issue with the auction requires approval and even after setting always allowing privacy settings I was still getting this message preventing my actions from running properly.

In the end I had to modify my schema to indicate that the action was safe from within the schema using “x-openai-isConsequential”: false

7

Yes. That works. Figured that one out long ago. Anyway. Thanks for letting us know. :slight_smile:
And you can for example use cloudflares wrangler’s to deploy functions/make the API calls work again, etc.