Zero Data Retention
Zero data retention is an arrangement in which a provider processes a request and stores nothing from it once the response has been returned.
It has become a standard procurement ask for enterprise AI, and it is narrower than it sounds. It usually concerns what the model provider keeps after serving a request, rather than what the platform a company is actually using stores. A platform has to retain a great deal to be useful: the context layer, the audit trail and the work in progress all persist by design.
The useful version of the question is therefore layered. What does the model provider keep, what does the platform keep, for how long, and is any of it used for training? Those have different answers, and a vendor that has answered only the first has answered the easiest one.