CrushBtc

Sharing a simple way to detect whether GPT has been downgraded

1️⃣ Ask it about something that happened recently. For example: “Without going online, tell me who the prime minister of Japan is.” If it doesn’t give the current one, it’s basically routing you to 5.6 luna.

2️⃣ I found a fingerprinting website that can be used to check. I tested it three times, and it’s pretty safe to say it’s routing to luna.

That’s not the problem—the main issue is that it’s just too dumb.

Detection link:

This is really bad, guys. Whatever you do, don’t use ClubSIM to link a Google account. I used this phone number to link my Google account, but Google prompted me to verify it by sending an SMS. There’s no way to send text messages from mainland China, and roaming doesn’t work either. Otherwise, I won’t be able to use Google. Looks like I’ll have to find time to take a trip to Hong Kong. I’ve also signed up for a lot of services with this Google account, such as GPT PLUS, 48TEAM, and some servers.

This is the kind of sense of responsibility and security awareness a major exchange should have. I hope BN keeps up the good work and helps make the entire industry better.

Based on what I’ve observed, this is probably the main reason for the last wave of account bans. These BIN ranges are mostly used for virtual cards, and Chinese users are concentrated there. There have also been a lot of chargebacks on these BIN ranges, so OpenAI couldn’t recover the payments and banned the accounts. You can review whether everyone who was banned was using these BIN ranges.

As for reverse proxies, they’ve always been tolerated. IP quality generally affects model performance, not whether your account gets banned.

I can finally buy higher-tier AWS Lightsail servers today. For a while, I could only get the 5u and 7u plans. Whenever I tried to launch anything higher, it just wouldn’t go through. I couldn’t find any tutorials online, and even GPT said there was no way around it. Then I submitted an appeal, and to my surprise, it was approved. I tested plans 24 and up this morning, and they all work. Anyone who knows their stuff—are these server plans actually worth it?

Oh, and there’s another trick: the PR rule of shifting the blame elsewhere. Any statements unfavorable to BN are the work of competitors; any public opinion unfavorable to BN is being manipulated by someone.

NVIDIA’s CEO is suspected of having an affair with a foreign female spy and is reportedly under investigation by NVIDIA’s board. NVIDIA is already down 3% on the 1H chart. Sure enough, men go bad once they get rich.

This incident is seriously problematic. No matter what went wrong, it’s always the users’ fault, never BN’s—not 1011, and certainly not 1031. BN is never wrong; it’s always the users’ fault, or some objective circumstance, or someone else’s fault. Normally, this incident shouldn’t have involved a UID collision. BN is a major brand, and across multiple sites, UIDs should be unique. Instead of blaming the users and making excuses after the fact, they should take responsibility.

I really can't take this anymore. That second confirmation step Yijie mentioned—how prominent is it? Has anyone ever made a cross-platform transfer? I'm genuinely curious what this second confirmation step looks like. I think this is a design issue: BN promotes UID transfers, and users are used to them. Even when transferring between different platforms, people will default to using a UID, right? But if UIDs collide across platforms, that's clearly a design issue. All I can say is that this isn't solely the user's fault.

This is basically a speedrun to getting your account burned, guys—check it out. Just be sure to use an alt account; OpenAI likes to settle old scores. But this idea is seriously brilliant.

Additional research: Are the credit cards linked to the banned accounts in any of these BIN ranges: 544015, 546400, 543502, 543799, or 512998? I suspect the virtual cards all use these BINs, and that the BINs ran into trouble, causing the accounts to be temporarily suspended. However, it looks like some people have had their accounts unblocked. Link:

No worries, Chinese people have their own way of doing things 🫡

Additional note: The account mentioned above is still active, and there’s also a 48team that’s still active, so the risk-control model is likely the problem.

Vote on why ChatGPT accounts are getting banned—let’s figure out what’s really going on together.

There was no email notification for this round of account bans. After logging in, all I saw was that the accounts didn’t exist. The reason is still a mystery: I have two Pro100 accounts, and my IP changes frequently. One is in a non-US region and one is in the US region, and I used both with sub2. One claimed the API quota and used about half of it on sub2. So my take is that this was another false positive from the risk-control model. If you’ve been banned, don’t panic—just appeal and wait for compensation.

Appeal portal:

Breaking: OpenAI may be banning accounts en masse. Plus, Pro 100, and Pro 200 accounts seem to be affected across all regions. The reason for the bans is still unclear, but it probably has nothing to do with reverse proxies. Check if your account is still active.

Anyone who has claimed this credit before should use it with caution. I’ve seen a lot of people on L site whose GPT accounts were banned. Generally speaking, OpenAI’s API platform has very strict risk controls. If you get banned, your account will be banned too. Codex, on the other hand, is very stable when accessed through a reverse proxy. If you’re connecting it to an external agent framework, definitely don’t use the API credits.

One more thing: when things are quiet, share some genuinely useful, substantive content—make sure it provides value.