I’m facing the exact same issue, but now I have a banner that says:
“Suspicious activity detected.
It looks like someone else may be using your ChatGPT account. Please secure your account to regain access to all features. Learn more.”
I’ve done everything in those instructions. I even reluctantly turned off the VPN and gave them my real IP. But nothing helped. The message keeps popping up, I’m restricted to the 5.5 mini model, and there’s nothing I can do about it. I contacted tech support, but I’ve heard nothing back. This is a complete false positive—I’ve never shared my account with anyone. My sessions are limited to my own personal devices. I haven’t been able to use the account I paid for for five days now. I’m a paying user, yet I’m not getting what I paid for. It’s absurd. Someone needs to fix this error.
Looks like a crack down on account sharing possible geo pricing abuse. If you used vpn before and your ip/countries changed a lot this could be it like say you got vpn on pc and nothing on mobile it looks like 2 different people 2 different countries, best give your real ip
So what to do now?
It started exactly the way it’s described in this thread: a silent model downgrade. For a while, clearing the cache and re-logging into the account helped. At some point, that stopped working entirely. I sent technical support all the details regarding my device, sessions, and so on. The situation has now escalated beyond what was originally described in this thread and become even more absurd; they’re telling me my account is “insecure,” even though the consequences are the same as those discussed here.
Essentially, it’s a downgrade to the 5.5 mini model for no apparent reason. I have two-factor authentication enabled; I use the web version 80–90% of the time, occasionally use Codex, and use the mobile app somewhat more frequently. The devices I use never change, yet the problem persists. At first, it happened silently; now, at least, a “security” banner has appeared. But the banner doesn’t make me feel any better, since the problem remains unresolved.
Perhaps others in this thread are facing the same underlying issue, but simply aren’t seeing the banner? After all, nothing showed up for me for four days either—it only started appearing yesterday 
Wait or get a new account/ip/hardware and software identity fingerprint and even if you didn’t do anything wrong before on purpose just be careful
Thanks for looking into this.
My affected account is under Support Case #13570204. This is the same case associated with my earlier reports in this thread as yuanzhe007.
The case already contains the HAR, screenshots, timestamps, and backend routing evidence. One reproducible request was:
x-oai-request-id: 00d2b1da-6a30-4f4a-bc43-ca379c384614
Timestamp: Sun, 23 Aug 2026 13:27:20 GMT
Requested: gpt-5-6-thinking / thinking_effort: extended
Resolved: gpt-5-5-mini
No usage-limit or fallback notification was shown.
Thank you for reviewing the case.
As an update, the last time I observed the routing issue was at approximately 13:00 on August 26, 2026 (Asia/Shanghai), and the first time I confirmed that routing had returned to normal was at approximately 17:30 the same day.
I did not record exact timestamps for these checks, so these times are approximate. The actual recovery therefore appears to have occurred sometime between approximately 13:00 and 17:30 (Asia/Shanghai).
I created a new account and purchased a fresh $20 subscription. The error message is gone, but now the system silently reroutes me to the basic model—the one without reasoning capabilities—without even acknowledging it. When using “High Sol 5.6,” it spends zero seconds thinking about a complex astrophysics problem; there’s no “Thought for X minutes” message—just an instant response. And that response is riddled with errors that a reasoning model would never make. So, even creating a brand-new, clean account—using a fresh, private home IP—didn’t help; the issue described in this thread persists even on the new account.
I really hope it gets fixed. But I can’t even provide a ticket number because no one is responding to my support requests. Maybe the OpenAI developers will see this and at least let me know that my inquiry is being looked into…
Oh wow sorry to hear so you saying fresh IP. Is it like proper IP no VPN etc like calling up broadband provider or keeping your phone off for some time to force reset? And re device software identify is it like browser or something? Brave is good. And same for card. Sorry it didn’t work for you
The team has already deployed a fix that has helped some affected users and they are still looking into it.
Please follow up in the topic linked above!