Hi team, I am speaking with a potential client who has some questions: 1. Is any customer data used to train models? 2. What is stored by default, and for how long? 3. Can we run with no retained call data after the call? 4. What still remains on Vapi infrastructure even then? 5. Can you send the formal security and legal documents? Thanks

Hi, Thank you for reaching out!

Hi, I couldn't find answers to the following in the docs: 1. Can you confirm whether call audio, transcripts, structured outputs, metadata, prompts, and user responses are used to train, fine-tune, improve, or evaluate any public, shared, or third-party AI models? 2. Your terms say customer content is not shared with third-party services for their independent model training purposes. Can you confirm this applies to all providers involved in the voice pipeline, including LLM, speech-to-text, text-to-speech, and analytics providers? 3. Can we fully opt out of Vapi using customer content for service improvement or call-improvement features?

Hi there, 1. Training / fine-tuning No – call audio, transcripts, prompts, or any other customer content are not used to train, fine-tune, or evaluate public, shared, or third-party AI models. 2. Pipeline-provider guarantees Our Terms state that we do not share your content with third-party services for their independent model-training purposes; however, each provider’s own terms govern anything they do with the data once we hand it to them for real-time processing. If you need contractual assurances, review each provider’s policy or use your own BYO keys. 3. Opting out of Vapi service-improvement use There is no customer-visible toggle, but you can enable Zero Data Retention or HIPAA mode (both remove persistent storage). Full detail and subprocessors are published at [https://security.vapi.ai](https://security.vapi.ai) – please review that page and perform any additional due diligence you require. Best, Vapi Support