Data privacy

V12 uses multiple models during a run. Settings → Privacy controls which kinds V12 is allowed to use.

Zero data retention (ZDR) restricts V12 to providers that contractually guarantee your code and prompts are never stored or used for training. The strongest models support ZDR, so most runs are unaffected. The available model pool is smaller, which may occasionally change which model handles a given task. Pricing is the same either way.

Enable ZDR if your organization requires that source code is never retained by a third party. Otherwise leave it off.

Workspace memory is V12’s record of your team’s decisions. When it’s relevant to a task, excerpts are included in the prompt sent to the model, so the provider processes them like any other prompt content. ZDR governs what the provider may retain, store, or train on—not whether it receives the prompt. With ZDR on, those excerpts are covered by the same no-retention, no-training policy as your code; with ZDR off, they follow the provider’s standard contract.