Had my account banned 6 months ago. Created a new account and have continued with no issues since

Lucky you.

I know it’s not ideal, but why not give the main competitor a go? I have great experience of using OpenAI Codex, and frequently switch between it and Claude on a whim, often playing them off against each other to review one another’s PRs.

Don't have a solution but just to add that I'm the same boat, and I've seen a few others. Something trips their anomaly detection and they don't care about it because it's presumably a low false positive rate.

For what it's worth I can just pay for tokens through other providers proxying their API. Still sucks because you end up paying much more.

Thy're probably using their own AI to find violations. Go figure.

Why not using some other coding IDE with OpenRouter? There you can also use the models. You can even switch models easily depending on your needs and spent.

There are OpenSource versions of CodeX.

For example:

* OpenCode: https://medium.com/codex/kickstart-opencode-with-openrouter-...

* CodeX with OpenRouter: https://openrouter.ai/docs/cookbook/coding-agents/codex-cli

Use grok build. They've got some pretty cool things going

Well at least you didn't submit an ID and selfie yet so you could make a new account right, the future is bright.

The part that would bother me most is not the ban itself, but having no idea what to change. “Policy violation” could mean VPN, billing risk, location, or actual usage.

Just generate a virtual credit card?

There's just a lot of fraud. My bank killed my debit card for "fraud" in my attempts to pay my anthropic bill which triggered fraud alerts, I confirmed the charge was legitimate and it got flagged again. Three times until they just deleted my card.

I too was banned. Two key details: 1. I was unbanned via their appeal form, but 2. I had created a new account in the meantime which I’ve used since.

If you need a debit card, use Mercury https://www.mercury.com. You can create as many as you want. I use a unique card for each subscription. (If you’re thinking "oh, they’re like privacy.com," you’re mistaken. Privacy.com gets blocked by merchants; mercury works. No idea why.)

Since you’re already throwing around $120, I suspect Mercury’s sign up fee won’t be a problem. But if you want to bypass it, you can start an LLC and use your EIN during account setup for free Mercury access. I did it for about $20 total and have been using Mercury for about 5 years. Best bank ever, and I’ll happily shill them all day every day till more people try them out.

In short, just dodge the ban. I used the same name and address on both my accounts, so as long as you use a different card you have a decent chance of bypassing the ban.

Strange, I managed to sign up with an email alias and pretty much always have my proton vpn turned on, never had any issues with Claude code or web - maybe I’m about to? :D

Been thinking of switching to a multi model harness anyway.

Might not be a very popular advice or even what you're looking for, but I think Anthropic banning you (or any user for that matter) is a blessing in disguise. The sheer amount of compute resources consumed by their high reasoning "pondering..." and "bloviating..." tokens isn't sustainable at scale. Eventually, they must ban everyone but those with deep and infinite pockets in order for this model to be sustainable and turn revenues.

Computers and LLMs are great at automation of low-level human cognitive tasks like memory, decisions and loops, etc. but struggle enormously with high cognitive tasks like reasoning, deep logic, nuance, etc. Not that it can't be done (Claude platform is proof that it can) - but the cost and scaling advantage in this realm belongs to the human brain, not the LLM.

Opencode, w. openrouter?

Kiro.dev, w. cerebras?

Aider, w. groq (with a Q)?

Just use one of the alternatives, there are so many. I recommend DeepSeek, GLM/Z-AI or Qwen. They are cheaper and offer competitive performance to Anthropic.

Very simple, just don't use it and switch to another model (e.g. a Chinese one).

Don’t worry about it, Anthropic is banned from releasing new models anyway.

Try out some of the competitors, they are really good these days.

Deepseek models seem as good and are like 20x cheaper for me.

Probably they think you’re using a personal account for your company, they don’t want that as they heavily (?) subsidize personal usage.

This brings up so many issues imo.

I hate the idea of government throwing its weight around based on personal vendettas (in the case of this Fable debacle), so it's clear that if this tech is going to be foundation-level important to the economy going forward, we need some sort of laws guaranteeing our access to it.

But authoritarians in government aren't the only party we need to be concerned about. As shown by this post, the actual model companies themselves may have too much centralized power already.

Given all software development has essentially moved to AI-first, an authoritarian-minded Anthropic/OpenAI employee is currently able to pick winners in the economy by granting/withholding access to certain groups. That is the type of thing I think needs to be regulated, not some trivial cyber security abilities in the actual models themselves.

The Google-style B2C blanket ban with zero customer support approach isn't going to cut it if the models continue progressing at this rate and the lead ever widens with open source (which it likely will at some point).

Reminds me of Google banning first time small business owners with no explanation and basically forcing them to shut down their businesses over night because without Google they couldn't find any customers. Happens way more often than people realize. Usually, the cause is something dumb like not having a link to a privacy policy clearly shown on the main page of their website but Google never explains this.

I've been using both codex and claude for the past month and if anthropic acted up like that I'd switch fully to codex in a heartbeat.

I try not to depend on these AI tools, not having them is inconvenient but I still write every line of code myself, so the speed I work at remains always the same.

I also never keep plans running for months on end. I pause them most of the time relying on free usage and resume them during busy hours when I run out of free usage. writing the code myself is also more fun tbh.

>I couldn't even move to real coding in 1 hour

That sounds... concerning? I would try to manage the context of the work you're doing yourself and only consult it for certain functions.

I don't think you need the very best from Anthropic if you follow these rules, you might also save quite a bit of money.

Sorry If this sounds like a lecture I just wanna give tips to continue your work. Best of Luck in getting your account back!

Is this the new “google banned me for no reason”?

Googles customer relations were “AI” before LLMs.

In all seriousness, between DeepSeek-V4-Pro, Kimi K2.6 and now K2.7-Code, Xiaomi's MiMo-2.5-Pro, and Qwen3.7 Max, I haven't touched Claude for any sort of programming-related task in months.

I still lean on Claude for research/chat questions that require going out to the world to get the answer competently, but that's just laziness, and all of their competitors can do that, too.

I don't use OpenAI or Gemini, either. The Chinese models are just that good. If all three of the US majors banned me, I think that'd be just fine.

At the end of the day, the only guaranteed way to retain access to LLMs is to run an open model on your own hw. Which at this time, it's often inferior but may change in the future.

Also, I'll never be grateful enough for all the horror stories about Google accounts getting randomly banned - they were what finally pushed me to make a similar move.

It's amazing that companies still can't handle the concept of somebody moving to a different country but still using a credit card from their original country.

I spent two days on the phone with Microsoft last year trying to find a way to pay them $700 to renew my lapsed Visual Studio subscription. I live in France and want to pay with a US card (or UK card or even a French card at this point), but because of some combination of physical location, store location, account location, card country, and vpn that I had on the first time I tried the process, the system is in a state where it can never again process a payment for me.

There are like 5 or 6 companies where I'm in this state. You get exactly one try to guess the magic combination of all those things above to get the backend to sail you through smoothly. If you blow it, it'll write everything down and refuse to let you change anything, then drop you into an infinite loop telling you to just change [store|country|card country|hairstyle] and sending you back to the beginning.

I mean sure, it's probably saving me a couple grand a year in services that I wish I could get working. But it baffles me that those companies don't want that money for themselves.

The permanent solution to this is: *Run a local model.*

Why the fuck are you renting your hammer, carpenter, from assholes who will capriciously take it away from you when you need it most?!

Sorry, no, but I’ve been using the latest DeepSeek model for personal stuff and it’s pretty good, way cheaper, and can be used in Claude Code directly.