"Our systems are thinking a bit more about this request before responding..." warning message

ChatGPT
,
1

Hello

Keep seeing this warning once in a while at top of the response being though through:

“Our systems are thinking a bit more about this request before responding. You can retry with a faster model for a quicker response, though it may be less capable of handling complex requests. Learn more”

Clicking learn more says something about hacking or biological related issues

None of my queries are related to that I’m just using it to research a small project

Is this capacity related as I never see this in Work/Codex (at worst I saw capacity and after a few retries taking very little time it resumes)

2

It is probably an account flag\restriction or there might be some flagged chats in your account that the system is flagging.
There is no way to know if you have any of it.
There is no way to know how long it will last.
I have the same problem, they don’t do anything about it.

3

Fair enough, not saying I disagree, I only have one account, I did try searching online but not much mentions, maybe you right I accidentally doing something wrong, but regardless thought is a good idea to ask if others have similar issue

My only thinking is hidden throttling since if I wait long enough I eventually get a response but I don’t know is it genuinely throttled or slowed or what. I guess one way to test it is get liked a detailed research task prompt and run it a few times keeping out for run times and if this warning pops up

But I agree could be false positive like once I got cyber access reply but I was analyzing my extension which is no way related to hacking (it’s for tracking ChatGPT message count and thinking time)

4

It is not only you “accidentally doing something wrong”.
If the model even without your instructions reads something about cyber\chemistry\biology, it could falsely trigger the additional security checks.
If it was triggered in that chat and that information\sources stays linked in that chat contaminating it.
In a new chat the system could read\see the content of the contaminated chats, and again falsely trigger the system in the new chat, thus contaminating the new chat as well.
As the system is triggered again and again, the probability of an account flag\restrictions increases and the problem worsens.
This is purely speculation, I don’t actually know how the safety systems works, this is how I suspect it works based on my experience.

5

I’m seeing something very similar.

In my case, it started after several chats where ChatGPT was reading/analyzing my own GitHub project. A few of those conversations unexpectedly triggered cybersecurity-related restrictions, including responses saying that some actions would require cybersecurity authorization/licensing, even though I was just working on my own software project.

Since around that time, the behavior of my account has changed noticeably. I now get the “Our systems are thinking a bit more…” message repeatedly across different chats, including normal project work that has nothing to do with cybersecurity, biology, chemistry, etc.

The bigger problem for me is not just that warning. New or existing chats sometimes fail to load properly, generation appears to stall completely, and I occasionally have to refresh the page. During the delay I don’t see the normal activity/progress information either — it can look like nothing is happening at all until the request eventually completes or times out.

On some of the worst cases I have waited roughly 30–37 minutes before anything appeared.

I’ve been using the same account and essentially the same project workflow for more than four months without this behavior. It only became a serious problem in the last few days.

I can’t say whether this is an account flag, hidden throttling, an additional safety-review layer, or simply a backend/UI regression. But the timing is interesting because the problems became much more frequent after those false-positive cybersecurity detections.

If OpenAI staff see this thread, it would be really useful to know whether the “thinking a bit more” banner means additional safety processing, capacity/routing, throttling, or something else — and whether repeated false positives can affect unrelated conversations on the same account.

6

30-37 minutes is well above the reported 27 minute hard cut off limit I assume you are using Work or Codex otherwise it just doesn’t make sense

Yeah I noticed sometimes it overreact to specific wording but I still like to think it’s some hidden throttling when you ask it too much

7

I’m experiencing the same issue as well.

When this message appears and interrupts my work, I’ve sometimes been able to work around it by opening a new chat, pasting the link to the affected conversation, and asking ChatGPT to check what was happening there. In some cases, I can then continue the work normally in the new chat.

Still, this issue has become extremely frustrating. I’m simply working on my own projects and not doing anything that violates, abuses, or interferes with anyone else’s services.

I really hope OpenAI is aware of this issue and can investigate it, because having longer tasks suddenly interrupted like this makes it very difficult to work reliably.