Operators can expand or restrict ChatGPT's default behaviors, i.e., how it behaves absent other instructions, to the extent permitted by OpenAI's guidelines.
This clause shifts significant compliance responsibility onto API operators — if a developer's platform enables a prohibited use, the operator bears legal and contractual liability, not just OpenAI.
OpenAI's Usage Policy defines hard limits on what consumers and developers can request from ChatGPT, Sora, Codex, and the API — including absolute bans on generating CSAM, weapons-related content, and content designed to undermine AI safety systems. Violations can result in immediate, unilateral account suspension or termination, which for paying subscribers or businesses relying on API access represents a material service disruption risk. You can review the appeals process for enforcement actions at https://openai.com/transparency-and-content-moderation.