Realtime AI News
OpenAI Deactivates Accounts at Scale Without Warning, Hitting Cheap-Region and Shared Subscriptions
OpenAI deleted a large number of ChatGPT accounts without email notice on November 8, hitting cheap regional subscriptions, shared Team plans, free API accounts and even $200 Pro users. The sweep follows device and payment-card limits introduced on November 2 and precedes a November 20 policy requiring official authorization for third-party API gateways.

On November 8, a wave of ChatGPT users reported that their accounts had been deleted outright with no email notice, leaving only one line of error at login: "You do not have an account because it has been deleted or deactivated." According to a report republished by PEdaily from the outlet Xinzhiyuan, the sweep hit users on cheaper regional subscriptions such as the Philippines and Turkey, low-cost Team plans, and accounts routed through reverse-proxy relays.
The cleanup was not limited to free or discount tiers. Free API accounts, shared team subscriptions and the $200-a-month Pro plan were all caught up, according to the report, along with Pro 20x, Plus and student-verified accounts. It also cites users who had renewed for four or five months, and one who said a five-year Pro subscription was disabled with no warning, no email and no stated reason.
Complaints piled up in the X replies of Tibo, the former head of OpenAI Codex. Developers described work interrupted mid-project, and at least one said years of conversation history and personal memory were locked in the cloud with no way to retrieve them; others said appeals against the deactivations went nowhere.
OpenAI has not published a general explanation. The report notes that its status page had flagged elevated error rates for some US commercial Mastercard cards, listing BIN prefixes 544015, 546400, 543502, 543799 and 512998, with some cards starting 513989 also affected. OpenAI framed that as a payment-rail problem and did not mention account bans.
The report places the bans inside a broader tightening of the paid ecosystem. On November 2, it says, OpenAI published a document on payment-card and device limits that brings Go, Plus and the Pro 100, Pro 200 and Pro 500 tiers under hard quota controls. Linking too many paid subscriptions to one card or one device now triggers a "Too many subscriptions associated with this card or device" error at checkout, and once "Device limit reached" appears, signing out of the old account or buying extra Credits does not free the slot.
A second change targets access routes. From November 20, a new OpenAI usage policy requires every third-party API gateway and model-routing intermediary to hold written official authorization, which would rule out unofficial relay services on both policy and interface grounds.
A third change concerns billing. Starting in December, ChatGPT and Codex usage on the Plus, Pro and Business plans will draw on a single shared quota measured in tokens, removing the option of rotating through extra accounts to top up capacity.
For individual developers and small teams, the practical risk is continuity of data and workflow: conversations and Codex projects live in the cloud, and an account that disappears takes them along. The report suggests filing an appeal through OpenAI's official form with the account details and the Request ID, while the safer habit is to sync important sessions and projects to local storage.
What to watch next: whether OpenAI issues a formal statement and a workable appeal path, how the November 20 gateway authorization rule is enforced, and how usable Codex capacity looks once the shared quota takes effect in December. Those details will show whether this is a cleanup of subscription arbitrage or something that reaches ordinary paying users.
Why it matters
Developers relying on cheap regional plans or unofficial relays face immediate loss of accounts, history and Codex projects, and should localize critical work now. The larger signal is that OpenAI is closing subscription arbitrage through payment, device, gateway and billing controls all at once.
Nearby Updates
All11/08, 14:54
DeepSeek Nears OpenAI's Cyber Model on Vals Test at About $3 a Run
Vals AI's CyberOffense leaderboard shows an open-weight DeepSeek model finishing close behind OpenAI's cyber model on tests where agents attack simulated networks, at roughly $3 per run. The result narrows a gap that had been assumed to separate restricted frontier systems from freely downloadable ones.
11/08, 10:50
Airbnb to add an AI agent for trip planning
Airbnb plans to add an AI agent that helps users plan and organize their trips, according to a YugaTech report. The move would push the short-term rental platform beyond listing search and booking into the earlier stage of travel decision-making.
11/08, 08:22
Alibaba's Qoder opens up a mystery coding model called Qoduck for free
Alibaba's coding tool Qoder has made a model called Qoduck free to use, according to Startup Fortune. The report describes Qoduck as a mystery coding model, suggesting that details about its origins and specifications remain limited even as access opens up.
11/08, 07:43
OpenAI's GPT-6 Cyber Finds 81 Firefox Flaws as AI Reshapes Security Work
OpenAI's GPT-6 Cyber, a model built for cybersecurity work, has reportedly identified 81 security flaws in Firefox. The case is being read as a marker of AI moving from code assistance into vulnerability discovery, and it sharpens the debate over how much offensive capability a defensive tool should carry.