The moderation product is for ensuring the safety of the AI inputs and outputs you run on the API. It is “no additional charge” and meant for that filtering scope, loosely enforced.
API organizations have an “on” switch. If you do not have a credit-purchase-funded, provisioned API organization account that has been unlocked and set up to make the underlying paid API calls to all models and endpoints, even the models endpoint to list which AI models are available, then moderation API calls will also not succeed.
Prepaid API organizations will also be turned back off if the organization’s credit balance goes negative.
So you are correct: a credit purchase is expected to use the API. Some new accounts that go through the “start building” wizard will have some free use of a mini model suggested during that onboarding that doesn’t immediately require credit purchase to make some test language generative calls.