Overview

OpenAI uses automated systems and user reports to identify content that may be harmful or violate its Usage Policies. Depending on the situation, ChatGPT may limit a response, restrict sharing, end a conversation, or temporarily suspend account access. OpenAI may also take separate action on an account or GPT.

These safeguards build on the safety and alignment testing and mitigation OpenAI conducts before releasing its models.

Identifying potentially harmful content

OpenAI uses automated moderation tools to identify potentially harmful content in prompts, responses, uploads, and GPTs.

ChatGPT can take certain safety actions automatically, without a person reviewing the conversation first. When someone reports content, OpenAI uses automated systems and trained reviewers to assess that report.

Actions in ChatGPT and ChatGPT Images

When content may violate OpenAI’s policies or raises a safety concern, ChatGPT may:

• Warn you that your content may violate OpenAI's Usage Policies.

• Decline to respond to a prompt.

• Prevent a conversation from being shared.

• End a conversation that raises a serious concern about violence or harm to others.

• Temporarily suspend your account if further conversations raise serious concerns about violence or harm to others after repeated conversation endings.

OpenAI may separately permanently close an account for serious or repeated policy violations.

Ended conversations

If ChatGPT ends a conversation, it no longer appears in your chat history or search results. You can start a new chat unless your account is suspended or otherwise restricted.

Temporary account suspensions

If ChatGPT temporarily suspends your account, you are signed out immediately and cannot sign in or use ChatGPT while the suspension is active. A temporary suspension is different from permanent account closure.

For more information about these actions, see: Why ChatGPT ended your conversation or temporarily suspended your account over concerns about violence.

Permanent account closure

OpenAI may permanently close an account for serious or repeated violations of its policies. This action is separate from a conversation ending or a temporary account suspension.

For information about permanent account closure and available account-review options, see: OpenAI account closures and safety suspensions.

Actions involving GPTs

OpenAI also uses automated tools to identify potentially problematic content associated with a GPT. If a GPT may violate OpenAI's policies, OpenAI may prevent it from being distributed.

The GPT's builder can update the GPT or API configuration to address the issue. If an in-product appeal is available, the builder can use it to request review. In especially serious cases, OpenAI may also permanently close the builder's account.

Reporting problematic content

You can report content that you believe may violate OpenAI's policies or otherwise present a safety concern. OpenAI uses automated systems and trained reviewers to assess submitted reports.

• For content directly in ChatGPT, see: Reporting content in ChatGPT and OpenAI platforms.

• For a shared ChatGPT conversation, see: Reporting harmful or illegal content in a shared link.

• For a GPT, see: Reporting a GPT.

If someone may be in immediate danger, contact local emergency services. Reporting content to OpenAI is not a substitute for emergency assistance.