all account deactivated all the sudden

[villa1]

Summary

Around 10 accounts were suddenly marked as deactivated within one day.

Most accounts had normal or very low usage, and no clear warning or explanation was provided.

Questions

• Is this related to a recent security or anti-abuse update?
• Was this triggered automatically?

Any clarification would be appreciated.

[NN1985]

I'm having the same issue. Sometimes accounts are all being "deactivated" the same day I log back in again. It's making the experience unusable. These are all legitimate paid-for accounts. I was forced into SMS authentication for each of them, which is inconvenient since it's max 2 accounts per phone number. But I can't understand why I'm constantly being logged out of each account. Is anyone else facing this?

I didn't have this issue for the previous months using multiple accounts and codex-LB until recently.

[haithium]

Same here - accounts die even mid-session

[haithium]

...which is inconvenient since it's max 2 accounts per phone...

Small correction - OpenAI says it's 3 times per phone number, not 2. https://help.openai.com/en/articles/8983031

[NN1985]

Same here - accounts die even mid-session

Did this behavior start for you recently? I had a nice little fleet of 6 Plus accounts that worked all the time without issue until recently.

[gabrielnogueiracsc]

same here! it started this week.

[haithium]

Did this behavior start for you recently?

Mine were fine until recently too

[NN1985]

Something appears to have changed in how token expirations are handled on the backend, whether through routine operational changes or the introduction of new security heuristics that are mass-deactivating tokens.

I'm looping in @Soju06 to flag this. If this behavior persists or escalates, it could have a significant impact on the project's normal operations and may require mitigation efforts. I'd propose implementing some method of studying these "deactivations" (or forced token expirations) to better understand the trigger conditions and explore potential workarounds.

For context: these are paid accounts, fully within terms of service.

[0xbffff600]

For whatever reason I noticed it started to increase once I moved my codex-lb instance to my VPS. I am moving it back to residential IP to see if it happens less. They may be using IP logs to try and filter out bad actors. Edit: Not saying we are bad actors but our usage may trip flags. Collateral damage.

[NN1985]

For whatever reason I noticed it started to increase once I moved my codex-lb instance to my VPS. I am moving it back to residential IP to see if it happens less. They may be using IP logs to try and filter out bad actors. Edit: Not saying we are bad actors but our usage may trip flags. Collateral damage.

Data point: I'm on entirely legit residential Verizon ViOS IP blocks and encountering this issue.

[mahdimasumigit]

Is it possible that aggressive pool balancers trip a red flag in openai server? just an idea.
maybe lazy balancers solve the issue.

[bjspi]

Wondering which "routing strategy" would be the best to avoid account banning...? Any ideas? Maybe it's due to several incoming requests from the same IP to different GPT-Accounts seen by OpenAI, hence the block? Even though I'm unsure what should be against their ToS if you just buy/subscribe and use several accounts, because: Why not?

Wondering if it would make sense as default routing strategy to first burn limits of ONE account, than for the next account etc instead of trying to "balance across all accounts", because than my computer / same IP makes several requests to OpenAI under different Account IDs. Maybe it would be better to first only request from my IP to ONE account_id, if limit is reached, the next account etc....?

[neslot]

I think its just multiple IPs connected to multiple accounts, would be pretty easy for them to detect.