I’m trying to get to the bottom of some recurring “suspicious activity detected” warnings I’ve been seeing in ChatGPT.
When it first started happening, I followed all of the help.openai.com recommendations from the FAQ covering that particular warning. I’ve changed my password and enabled 2FA, which do seem to be reasonable things to do. But, here’s the thing: I’m still getting them, even after doing that. I suspect it might be VPN related. I use a VPN (NordVPN) regularly on all of my devices (gaming rig, tablet, cel, and laptop) and I will jump into a chatGPT session on any of them at any given time, when I need to check something that ChatGPT can help me with. The only time I’m not on a VPN is when I’m strictly mobile (AT&T is my carrier) and away from a wifi access point. That means my IP address changes pretty frequently—sometimes showing up as residential, sometimes from AT&T, and sometimes as datacenter traffic from wherever Nord slung me to. I’m wondering if that’s triggering some kind of automated flag in OpenAI’s system that thinks my account is compromised or being shared.
I’m a Plus subscriber; is anyone else experiencing this kind of thing? Is there an official word on how VPNs or rapidly changing IPs are handled on the backend? Any recommendations on how to avoid false positives without compromising security or access?
Thanks in advance.
Randy (concerned user who’d rather be writing sci-fi than debugging their AI access )
i have been experiencing this same issue for the past 25 hours, havent received ANY response at all from the help desk, only response i’ve received is from the chatgpt support mail, and i’ve gotten this:
Hi there,
Thank you for getting back to us.
We understand how frustrating this situation must be, especially after you’ve taken all the necessary steps to secure your account. We truly appreciate your patience and the detailed information you’ve provided.
The temporary downgrade on your account is a precautionary measure triggered by our system due to activity that resembles account sharing or unauthorized access patterns. While we recognize that you’ve only accessed your account from your own devices, these alerts are automated and designed to prioritize account security.
How to Resolve This:
If You Suspect Unauthorized Access:
If you believe your account has been compromised, we recommend following the steps in our guide on Securing Your OpenAI Account.
Ensure Your Account is Secure:
Please take a moment to review our Account Security Guide to ensure all best practices are in place.
Understand the Alert:
You can learn more about these alerts and what triggers them in our Suspicious Activity Alert guide
Wait for System Reassessment:
Once appropriate security measures are in place, our automated systems typically reassess and restore full access within a few hours to a day. If the issue persists beyond 24–48 hours, please reach out again so we can escalate the matter for further review.
We appreciate your patience and understanding as we work to keep your account secure.
so, basically their solution is to wait it out lol
yeah, the waiting part seems to be where I am at right now. I’ve received the message only a couple of times in the past several days; that is a definitely improvement over the past several weeks. Maybe it’s because I haven’t initiated any sessions from my cel in a while.
The annoying part for me is when the down grade in access happens after that warning pops up. It causes me to lose my current context; the system shifts me from 4o to 4o-mini. The first time it happened, I lost the context for a really productive editing session on my nascent novel, one that I’d been in-session on for most of the day. I may be anthropomorphising a bit, so caveat lector, but when you spend all day trying to get a resonant bit of prose going for a chapter in your novel, and you are using ChatGPT as a combination muse/editor, there is a definite rapport, or rhythm, or something that I seem to establish in the session, and the LLM locks in to it. If anybody doubted an LLM’s ability to pass the Turing test, I invite them to look at transcripts of my novel writing sessions. It is kind of jarring when that rapport is lost.
I don’t have a solution to this but just wanted to add some datapoints and also as an attempt to voice this issue louder:
I’m not even using a VPN. In fact, I tell my ChatGPT (sometimes I’m on my desktop, iPad, etc) and I’ve had ZERO issues since I signed up for the $20/monthly subscription until very recently.
I am NOT sharing my account with anyone. It is strictly me. In fact, I have certain codewords that I “trained” it that’s very specific to me that no one else should know about. In fact, it said “suspicious” when I used a very niche keyword I’ve recalled successfully multiple times in the past.
Tried changing my password. Enabled, then disabled, then reenabled 2FA. Doesn’t do **** unfortunately. What else do I do to “secure” my account…?
After 8 hours of waiting, it reverted back to the state I wanted, but it clearly “lost” a good chunk of short term memory. I spent about an hour reminding some key high level overviews about it. After maybe 10 messages, I got the “suspicious activity” flag again and immediately downgraded to mini, lol.
Very frustrating. I’ve seen this post pop up multiple times on the OpenAI forums, Reddit, and other sites and it clearly seems to be a recent and recurring issue. Cannot for the life of me, find a viable solution, nor get connected to support. No indication they seem to be aware of this issue or have any intention of addressing this.
Really contemplating seeking alternative options to ChatGPT. Such a shame cause it was providing everything I wanted but a paid subscription kind of defeats the purpose if it leads to this. Will try to narrow down some other data points (does it reset after 2 hours? 3 hours?) and report back if I find anything helpful…
So I have no confirmation if this works for everyone, but here’s what I did (in order):
My model is set to 4o. After relogging, clearing cookies, browser history, etc in a fresh window, I changed the model to 4.5 and kept talking to it until I almost hit the limit. Then I switched back to 4o and it never went back to ‘suspicious activity’ again since.
I also sent like 4 messages to support demanding they address this. They never actually responded but maybe they got tired of me and just manually deactivated it
I also sent a scathing (but still civil) email basically linking multiple threads complaining about this, and telling them they’re going to lose dedicated users. Again, no response but I don’t think it really hurt my cause.
More likely #1 did the trick. Incidentally enough, I deactivated my 2FA cause it was a pain in the *** so that “secure your account” msg is total BS.
Good luck though and sorry if that doesn’t work for you. Saying I was pissed going through this is an understatement and I can only imagine how everyone else feels.
Seems I’m late to the party but wanted to add I’m experiencing the exact same thing. I also use dedicated IP VPNs (Nord and SurfShark interchangeably) for work and am a Plus subscriber. I’ve completed all of the additional security steps (password reset, log out all devices, enable MFA) and reached out to support (received a standard response within an hour - they advised the suspicious activity message would go away within a day).
Over 24 hrs later I’m still seeing the ‘Suspicious Activity’ message after my first new prompt, even when disconnected from a VPN. It’s clear this isn’t really ‘security’ related because anyone with my account credentials would still have access to all of my project files, message history, etc. The only ‘security restriction’ they’ve put in place is blocking access to the more expensive models I’ve paid for.
Very disappointed with this support process. I prefer OpenAI’s current flagship models over other options (I also have Gemini’s standard paid tier) but I’ll have to switch over if the customer service is this unreliable. I can’t imagine any reason they wouldn’t be able to switch features back on as soon as the security steps have been completed - especially after I’ve contacted support to confirm like so many others here.
Update: the restrictions were finally removed from my account about 48 hours after they were applied. Clearing cookies and restarting my browser had no effect yesterday, but everything worked today after no additional changes on my end.
I had contacted support ~4 times via the chat option. They continued to insist the system would automatically restore access, which it appears to have done. I still think it’s a huge inconvenience to have all paid features revoked for 2 days out of the blue with no immediate recourse.
If the real concern is malicious activity, a password change + adding MFA + logging out of all devices should restore access immediately.
If the issue is account sharing, they should require MFA setup and add an option to pay for additional concurrent user to immediately remove restrictions.
A soft ban for two days because the algorithm is ticked off is just sloppy.
I’m also angry, i just want to just punch the security system in the face and tear it apart. I don’t have plus but i really want chatgpt 4o access back. i sent an angry email to chatgpt and told the ai support agent to sent a human agent
I will be getting a response from a human agent in a day or two
I am also experiencing issues. I use Opera’s VPN, which till now has worked without problems with ChatGPT.
And like you I use chatGPT primarily for assistance with sentence and paragraph revisions of my novel. The shift from 4o to 4o-mini really broke the flow.
As everyone else, I guess I will wait the recommended 48 hours.
On a side note, a power outage occurred two days ago, lasting seven hours. I am wondering if the internet connection has been unstable since then, causing the “suspicious activity” alert. Although I am not noticing a spike or drop in packets.
I think I know what caused this for me. I noticed that it happens when I log one day with my email and password and then log back six hours later though the google account.
Thats not what im paying for! Every time i use it a bit more it starts and takes a day or so to go away again. All the recommended thinks are not working, even if i do them all… @OpenAI_Support do something!!!
)