Contents

"One day my ChatGPT or Claude account just stopped working." For anyone who relies on these tools for work or development every day, that is a serious problem. What makes it worse is that you can get suspended without any bad intent, simply by breaking a rule you didn't know about, and that some people make things worse after a suspension (for example, by coming back on another account).

For this article we read the actual terms and help-center pages of OpenAI (ChatGPT, Codex) and Anthropic (Claude, Claude Code) from start to finish, and sorted out: (1) what gets accounts suspended, (2) where the gray areas begin, and (3) which channel to use to appeal. This is not a guide to evading detection; it is a practical guide to staying within the rules. We don't state anything more strongly than the clauses do, and where the companies say nothing, we say so.

ACCOUNT BAN PREVENTION · OpenAI / Anthropic

No bad intent needed: rule violations get accounts suspended

— Know what the terms prohibit, and stay ahead of it

🚫
Bypassing safeguards / jailbreaks
🤖
Unauthorized automation / data extraction
🔑
Sharing or reselling accounts/keys
🌐
Use from unsupported regions
👥
Returning on another account after a ban
⚠️
Handing your login to someone else's product

⚠️ If you're suspended, don't open a new account — use the official appeal

Versions read as of October 4, 2026: OpenAI Terms of Use (effective January 1, 2026) and Usage Policies (effective October 29, 2025); Anthropic Consumer Terms (effective October 8, 2025) and Usage Policy (effective September 15, 2025; a new version takes effect on November 12, 2026), plus each company's help center. Terms change, so always check the current originals. This is a general advisory; for your specific case, ask the provider.

1. The suspension reasons both companies officially list

Start with what each company says in its own help center about why it suspends accounts. Both terms also state that the company may suspend or terminate access if it believes you have violated them (OpenAI Terms of Use, "Termination and suspension"; Anthropic Consumer Terms, "Termination" — Anthropic explicitly says "without notice").

OpenAI (ChatGPT, Codex, API)

Source: Why was my OpenAI account deactivated?

  • Usage Policies violations (sexual content, child exploitation, violence and self-harm, scams, hate and spam, etc.)
  • Terms of Use breaches (circumventing security or access restrictions, sharing accounts or API keys inappropriately, etc.)
  • Temporary suspension when the account appears compromised
  • Not correcting usage after warnings
  • Not completing identity or age verification in time

Anthropic (Claude, Claude Code)

Source: Safeguards warnings and appeals

  • Repeated violations of the Usage Policy
  • Account creation from an unsupported location
  • Terms of Service violations

The Usage Policy says violations can lead Anthropic to "throttle, suspend, or terminate" access.

Two items are easy to miss: OpenAI's uncompleted identity or age verification and Anthropic's account creation from an unsupported location. Neither requires you to have "done something wrong." If ChatGPT asks you to verify, follow the steps and deadline shown in the app, as the same help page advises.

📅 Anthropic's Usage Policy moves to a new version on November 12, 2026

Anthropic announced a Usage Policy update on October 8, 2026, and the new version takes effect on November 12 (announcement, text of the new version). Until then, the version effective September 15, 2025 applies, and the sections of this article are based on it. According to the announcement, most changes clarify existing rules. These are the points that matter if you want to avoid account suspension.

  • Abuse of the models: sustained and needless abusive or cruel behavior toward the models becomes a prohibited use. The announcement says it applies only in extreme cases repeated with no discernible purpose, and not to ordinary frustration or pushback, dark creative themes, or model testing and research. The main enforcement remains Claude ending the conversation, as before.
  • Deceptive campaigns: rules that were scattered across the elections, fraud, privacy and disinformation sections are consolidated into a new section, Do Not Engage in Deceptive Campaigns or Artificial Activity. Political or commercial, it covers hiding who is behind content through fake accounts or personas, astroturfing, fake reviews and bots that claim to be human, as well as building tools for this.
  • Elections: the section is renamed Do Not Undermine Democratic Processes and narrowed to deceiving voters or disrupting elections. The blanket ban on personalized vote and campaign targeting is removed, but targeting that relies on deception or misuse of personal data stays prohibited under other sections.
  • Unsupported regions: the supported regions page now spells out how the policy is enforced. Besides people physically located in an unsupported region, entities incorporated or headquartered there, and entities majority-owned or controlled by persons or entities there, cannot use Claude even if their users are in a supported region.

The ban on circumventing a ban through a different account stays in the new version with the same wording (we checked both versions on October 12, 2026).

2. What not to do: concrete examples with the clause behind each

Legal language is abstract, so here it is translated into situations that actually come up, each with the clause it rests on. We don't claim more than the text says.

❌ Sharing one paid account with family or coworkers

Basis: OpenAI's Terms of Use (Registration) prohibit sharing your account credentials or making your account available to anyone else. Anthropic's Consumer Terms likewise prohibit sharing "Account login information" or API keys and making your account available to others. Get one account per person, or use an organizational plan such as Team.

❌ Driving the ChatGPT or claude.ai interface with a script to harvest answers in bulk

Basis: OpenAI lists automatically or programmatically extracting data or Output among prohibited uses. Anthropic prohibits accessing the Services "through automated or non-human means, whether through a bot, script, or otherwise," except via an Anthropic API key or where it explicitly permits it, and also prohibits crawling and scraping (Consumer Terms). If you want automation, use the API.

❌ Hunting for "unlock" prompts and trying them (jailbreaking)

Basis: OpenAI's Usage Policies prohibit circumventing its safeguards, and its Terms of Use prohibit bypassing protective measures or safety mitigations. Anthropic's Usage Policy prohibits intentionally bypassing guardrails to make the model produce harmful outputs ("e.g., jailbreaking or prompt injection") without prior authorization.

❌ Getting suspended, then opening a new account to keep going

Basis: Anthropic's Usage Policy explicitly prohibits circumventing a ban "through the use of a different account," including creating a new one, using an existing one, or giving access to someone who was banned. If you disagree with a suspension, use the appeal process in section 7.

❌ Using Claude from an unsupported country via a VPN that makes you look like you're elsewhere

Basis: Anthropic's help center says "only users physically located in one of our supported locations can create and use Claude accounts," and sign-up requires a phone number from a supported location (Verify your phone number). The Usage Policy also prohibits providing access to Claude in violation of the Supported Regions Policy. The list is at Supported countries.

❌ Collecting responses to train your own AI model

Basis: Anthropic's Usage Policy prohibits using inputs and outputs to train an AI model ("model scraping" or "model distillation") without prior authorization, and its Consumer Terms prohibit developing competing products or models. More in model distillation and the terms.

❌ Selling or renting out access to your paid account

Basis: OpenAI's Pro help page says "Sharing account credentials, reselling access and automatically extracting data are prohibited" (About ChatGPT Pro tiers). Anthropic's Consumer Terms also prohibit reselling the Services.

❌ Mass-creating accounts, or spreading violations across several accounts

Basis: Anthropic's Usage Policy prohibits using "automation in account creation" and coordinating "malicious activity across multiple accounts to avoid detection or circumvent product guardrails."

Sharing accounts and automating the chat interface are two practices to check before you start using the service. Before you let a browser extension or homemade script drive ChatGPT or claude.ai, check whether it goes through the API or is something the company explicitly allows. In a company, writing this into your AI usage guidelines prevents accidents.

3. Using your personal-plan login in other tools (February 2026 vs. now)

Can you use the credentials (OAuth token) you get by signing in with a personal Claude plan (Free/Pro/Max) in tools other than Anthropic's own? Anthropic's wording on this changed during 2026, so we'll take it period by period.

The wording in February 2026

According to The Register (February 20, 2026), Claude Code's legal page then said that using OAuth tokens obtained through Free, Pro or Max accounts in any other product, tool or service, "including the Agent SDK," was not permitted and violated the Consumer Terms. The same article reported an Anthropic engineer saying that third-party harnesses using Claude subscriptions are prohibited by the terms.

The wording as read on October 4, 2026

The "including the Agent SDK" sentence is gone from the same page (Legal and compliance). It now says advertised Pro and Max limits "assume ordinary, individual usage of Claude Code and the Agent SDK," and a help article (Use the Claude Agent SDK with your Claude plan) explains that Agent SDK, claude -p and third-party app usage draw from your subscription's usage limits.

So what is clearly prohibited now? As of October 4, 2026, the originals explicitly state the following.

  • Developers may not offer Claude.ai login in their own applications, or route requests through Free, Pro or Max credentials on behalf of their users (Legal and compliance).
  • Developers may not collect, store or intermediate Claude.ai credentials or session tokens; sign-in must complete through Anthropic's own flow (same page).
  • Developers building products or services, including with the Agent SDK, should use API keys from Claude Console or a supported cloud provider (same page).
  • The Consumer Terms prohibit automated access except via an API key or "where we otherwise explicitly permit it" (section 2).

From a user's point of view, the practical line is this. Running unmodified Claude Code or the Agent SDK on your own machine with your own account is a use Anthropic anticipates (see our Agent SDK guide). Giving your Claude login or token to a service someone else built, on the other hand, puts that service in prohibited territory and puts you at risk too. If you're building something for other people to use, build it on API keys (pay-as-you-go) from day one.

Anthropic has not published a list of which third-party tools fall under "where we otherwise explicitly permit it" (section 5). For unclear cases, the legal page itself points you to Anthropic sales.

4. Warnings, checks and restrictions that are not violations

Bans are easily confused with security checks and temporary feature restrictions. These are not findings of a violation. Before you panic and open another account, work out what is actually happening.

ChatGPT's "Suspicious Activity Alert"

Shown for logins from unexpected locations or devices, sudden spikes in activity, or more simultaneous logins than usual. OpenAI says these alerts "do not necessarily indicate wrongdoing." Fix: change your password, turn on two-factor authentication, log out of all devices. If it persists, turn off VPNs, proxies, extensions and automation tools and try from one trusted device and network (official help). Account sharing can also trigger it.

ChatGPT account warnings

OpenAI describes a warning as "a notice, not an immediate suspension." Reasons include generating harmful content, trying to bypass safety filters, automated spam, sharing API keys, and abnormal usage patterns. Continued violations after a warning may lead to further restrictions or permanent deactivation (official help).

Claude warnings and payment checks

Anthropic warns users when it believes their prompts violate the Usage Policy. You may also see a "Verify now" payment-method prompt at login; "Remind me later" postpones it for 24 hours (official help). If you're worried about unfamiliar logins, check your devices under Managing your active sessions.

Hitting a usage limit is not a suspension either. OpenAI's Pro help page says "A usage limit does not end your subscription or mean your account is restricted." For how the limits work, see our ChatGPT Pro comparison.

5. Are multiple accounts a violation? Ownership and use are different questions

“Can I have two paid accounts for myself?” and “Can I use a system to bypass limits?” are different questions. OpenAI officially explains how to use separate accounts. Anthropic also has help pages that account for multiple accounts. Neither establishes that every possible use is allowed. The following reflects a fresh review of the original documents on October 9, 2026.

OpenAI: official guidance on using separate accounts

Its guidance describes separating work and personal use, billing and plans, and says account switching supports all plans. It does not describe a blanket ban on holding multiple accounts.

The switcher's two-account limit is the number you can keep signed in within one session, not the total number you may own. Billing, history and memory are not merged.

This feature is for ChatGPT on the web. The help page we reviewed says it is not yet supported in Codex desktop or the native ChatGPT mobile apps. Organization-managed accounts may also be subject to administrator restrictions.

Official help: account switching

Anthropic: distinguish what the guidance actually covers

The officially documented switching feature covers a personal account and Team/Enterprise accounts linked to the same email address. That is a different situation from buying two personal Max subscriptions.

You can verify up to three accounts with the same phone number. This is a limit on the number of accounts that can be verified with that number, not permission to pool three subscriptions' allowances or switch between them automatically.

In the terms we reviewed, we found no clause imposing a blanket ban on one person having two personal paid accounts. The absence of that ban alone does not establish that your intended way of using them is permitted.

Official help: login and switching · Phone verification

What about manually switching to another paid account you own after reaching a limit? The public documents we reviewed did not settle whether that specific case is allowed. OpenAI's Terms of Use, “Using our Services”, do prohibit circumventing usage limits. It would be wrong to say there is no prohibition on bypassing limits, but that general clause alone also does not establish that every ordinary manual account switch is a violation.

Officially documented: separating work/personal and individual/organization accounts

OpenAI documents work/personal separation, and Anthropic documents switching between personal and Team or similar accounts. Owning multiple accounts yourself is different from handing your login details to someone else.

Ask about your specific case: manually moving to your other subscription after a limit

For neither company did we find a public document explicitly permitting this workflow. Before buying another subscription, ask support with the specific conditions: number of subscriptions, use by you alone, official apps and manual switching.

Explicitly prohibited: sharing with others, reselling and ban evasion

Both companies prohibit making your account available to others. Anthropic also explicitly prohibits evading a ban through another account. Paying extra, or sharing with family, does not exempt you.

Check separately from manual switching: tools that automatically route requests across subscriptions

A system that holds several sets of credentials and routes requests according to the remaining allowances cannot be judged solely by how many accounts you own. Check whether it involves unauthorized automated access, bypassing safeguards or providing access to third parties. This does not mean all authorized automation, such as official Claude Code, violates the terms.

OpenAI sets out a specific prohibition in the terms for apps integrating “Sign in with ChatGPT.” Apps must not enable or encourage creating multiple accounts, splitting usage or rotating accounts to circumvent usage limits (section 4 of the Sign in with ChatGPT Terms). These app-integration terms should not be applied wholesale to account switching in the ordinary ChatGPT interface.

For Anthropic, the sources are sections 2 and 3 of the Consumer Terms and the Usage Policy effective September 15, 2025. Among other things, the latter prohibits coordinating malicious activity across multiple accounts to avoid detection or circumvent guardrails; that is not a clause covering all ownership of multiple accounts. The newer Usage Policy page we retrieved displayed an effective date of November 12, 2026, so we treated that future version separately in this review.

Consider three things before buying another subscription: terms, cost and data. OpenAI's account switcher does not merge subscriptions, so paid subscriptions on separate accounts are billed separately. Before entering work information, check the email address, plan and organization you are currently using. For checks including use of data for training, see our ChatGPT and Codex data settings guide; for choosing a single subscription, see the Pro pricing comparison.

When contacting support, spell out the conditions: “I am one person with two personal-plan subscriptions, used only by me in the official apps. Am I allowed to manually sign in to the other account after one reaches its usage limit? I will not share or resell access or route requests automatically.” That separates whether you may own the accounts from whether the proposed use is allowed.

The documents we reviewed also did not settle the following questions.

  • Which third-party tools are “explicitly permitted”: there is no list (section 3).
  • Using a VPN within a supported region, or registering and paying from different countries: nothing says this alone leads to suspension. On ChatGPT, though, VPN use is among the items checked when troubleshooting alerts and feature restrictions (section 4).

6. A checklist to avoid bans

None of this is hard. Use the right plan, for the right purpose, honestly — that's the whole of it. Keep these eight points in mind.

✓

Read the terms and policies at least once: the four documents above, plus the help pages for the features you use.

✓

Use the plan that fits the job: products for other people and large-scale automation = API (pay-as-you-go); your own chats and development = personal plan.

✓

Never hand your login or token to someone else's service: sign in only through the official screens.

✓

Stay away from jailbreaks and prohibited content: don't try to break the safety filters.

✓

Don't share or resell accounts or API keys: if a key leaks, rotate it immediately.

✓

Turn on two-factor authentication and end sessions on devices you don't use: this prevents suspensions caused by account takeover.

✓

Finish identity, age and payment checks before the deadline: leaving them can get the account stopped.

✓

Act on warnings immediately: a warning is a chance to fix things. Don't keep doing the same thing.

7. If you're warned or suspended: how to appeal

If you have no idea why you were suspended, or believe it was a mistake, appeal through the official channel. Both companies document where to go.

OpenAI (ChatGPT, Codex, API)

  1. Appeal via the link in the suspension notification email (OpenAI says this routes your request faster).
  2. If you can't access that email, use the intake form at openai.com/form/appeal.
  3. If you lost access without receiving any email, contact support through the Help Center chat.
  4. Include: your User ID and Org ID, context about your usage, the date range of any suspected compromise, details of unrecognized charges if any (dates, amounts, last 4 digits of the card, etc.), and the steps you've already taken.

Sources: deactivation help and warning help (warnings can be appealed the same way). The Terms of Use also say that if you believe your account was suspended or terminated in error, you can appeal by contacting support.

Anthropic (Claude, Claude Code)

  1. Log in to claude.ai with the banned account (the form only opens when you're logged in).
  2. On the screen that appears (claude.ai/restricted), use "Submit an appeal." The Safeguards team investigates.
  3. The same screen lets you export your data and delete your account (how much you can export may be limited depending on the violation).
  4. If your own account is fine but an organization you belong to is on hold, click "Request a review" next to that organization.
  5. If you think a warning was a mistake, email usersafety@anthropic.com with the details of your situation and your account information.

Source: Safeguards warnings and appeals (updated July 9, 2026). The page also notes that response times are currently longer than normal.

Also know what not to do while appealing.

  • Don't open another account while you wait. At Anthropic, that in itself violates the Usage Policy (ban evasion).
  • Don't write out your full card number. OpenAI asks for the last 4 digits, the card brand and the issuing bank, not the full number.
  • Don't misstate the facts. Neither company publishes its review criteria, so there is no wording that guarantees success. Write what you know, as it happened.

Know the money side, too. Anthropic's Consumer Terms say that if access is terminated for violating the terms, you are not entitled to a refund of your subscription. OpenAI's Terms of Use say payments are non-refundable except where required by law.

Summary

The companies’ official reasons for suspension include terms violations and improper account sharing. This article is not based on statistics showing the breakdown of suspension reasons, so it does not claim which causes are most common.

Key takeaways

  • 🚫 Official suspension reasons: policy violations, terms violations (sharing, circumvention), suspected compromise, ignored warnings, incomplete verification, account creation from unsupported locations.
  • ⚠️ Personal-plan logins: using official tools or the Agent SDK yourself is anticipated; handing tokens to someone else's service or building them into products for others is prohibited.
  • 👥 Multiple accounts: OpenAI officially documents using separate accounts. Anthropic’s personal/organization switching and phone-verification limit do not authorize pooling multiple Max allowances. Ask about manually switching to another subscription after a limit, giving the specific conditions.
  • 📨 If suspended: OpenAI — the link in the email or openai.com/form/appeal; Anthropic — log in to claude.ai with the banned account and appeal. Don't open a new account.
  • 📅 From November 12: Anthropic's Usage Policy moves to a new version, adding a ban on sustained abuse of the models, consolidating the rules on deceptive campaigns and revising the elections section. The ban on evading a ban does not change.

Use “the right plan, for the right purpose, honestly” as your starting point. If a proposed use is unclear, describe the specific conditions to official support and ask for confirmation. See also our corporate AI usage guidelines, what not to put in prompts, model distillation and the terms, and the Claude Code vs. Codex comparison.

FAQ

Q. Can I get banned just for using it normally at work?

A. Normal use within the terms is nothing to lose sleep over. The reasons both companies list are policy violations, terms violations, suspected compromise, ignored warnings and incomplete verification. If you use it for work, though, avoid two things: sharing accounts and giving your login to outside tools.

Q. Is it OK to run the Agent SDK or my own scripts on a personal plan?

A. As of October 4, 2026, Anthropic's legal page says Pro and Max limits assume "ordinary, individual usage of Claude Code and the Agent SDK." In February 2026 the page said use in other tools "including the Agent SDK" was a violation, so be careful with older explainers. If you build a product for others, though, the page is explicit: use API keys and don't hold your users' Claude logins.

Q. Does using a VPN make a ban more likely?

A. We found nothing in either company's terms or help pages saying VPN use alone leads to suspension. On ChatGPT, logins from unexpected locations can trigger a "Suspicious Activity Alert," and turning off your VPN is one of the troubleshooting steps. Claude is only for people physically in supported regions, so using it from an unsupported country through a VPN breaks the rules.

Q. Is having two paid accounts a violation?

A. Holding multiple accounts alone cannot be treated as a blanket violation. OpenAI officially documents account switching across all plans. Anthropic’s phone-verification limit and personal/Team switching do not authorize every use of multiple personal Max subscriptions. This review did not settle the specific case of manually moving to another subscription you own after reaching a limit; ask with your exact conditions. Sharing, reselling and ban evasion are separate prohibited practices. See the comparison in section 5.

Q. Can I get my account back after a ban?

A. If you think it was a mistake, appeal: for OpenAI, via the link in the email or openai.com/form/appeal; for Anthropic, by logging in to claude.ai with the banned account and using the form shown there. Neither company publishes its review criteria or success rates. Don't open a new account while you wait.