ChatGPT Business: organization-wide instructions for every chat and Codex session

ChatGPT Feature requests
,
1

ChatGPT Business workspaces need organization-wide instructions similar to Claude Organization Instructions: hxxps://support.claude.com/en/articles/14546867-set-organization-instructions

Security/compliance teams need a short baseline prompt applied to every normal ChatGPT conversation in the workspace. The instructions should take priority over user personalization and should be centrally managed by workspace admins.

Today, the practical options are incomplete:

  • Personal Custom Instructions are user-managed and not enforceable.
  • Shared GPTs/Workspace Agents only apply when users choose that GPT/agent.
  • Codex uses AGENTS.md, which is good for coding workflows but separate from normal ChatGPT.
  • Workspace app/GPT/action controls are useful guardrails, but they do not replace baseline behavior instructions.

Requested capability:

  • Admin-managed workspace instructions for ChatGPT Business.
  • Applied automatically to every normal ChatGPT conversation in the workspace.
  • Also available to Codex sessions, or clearly mapped to managed Codex configuration / AGENTS.md.
  • Visible/auditable to admins, with version history.
  • Optional per-product scope: ChatGPT, Codex, GPTs/Agents, connectors/actions.
  • Clear precedence over user Custom Instructions.

Companies need a reliable place for short mandatory instructions such as:

  • do not reveal secrets, credentials, customer PII, or non-public data;
  • do not use unapproved tools, plugins, MCP servers, prompt templates, or memory sources;
  • do not perform AI-driven writes in privileged cloud/database/identity/production systems without approved workflow;
  • preserve audit context for security-sensitive work.

This would make ChatGPT Business much more usable for company-wide AI adoption without requiring every user to manually copy/paste personal instructions.

2

The strongest version of this is probably not one giant hidden system prompt. It should be an auditable instruction stack with scoped precedence.

I would separate four layers: non-negotiable organization safety constraints, product-specific defaults, repo/project instructions, and user preferences. The important part is not just applying the policy, but showing the effective stack when something behaves differently than expected.

For Codex specifically, an “effective instructions” view would help a lot: org policy version, AGENTS.md sources, skill instructions, MCP/tool policy, and user-level preferences. That makes conflicts debuggable and makes security teams more comfortable with rollout.