Tabnine
Tabnine says it never retains customer code beyond ephemeral inference and does not train its models on customer code, regardless of which Tabnine model is used.
Tabnine documents a no-train-no-retain policy for user code. Proprietary models are trained on permissively licensed open-source code; private fine-tuned models stay in the customer's private setup.
No opt-out is required for the stated baseline. Enterprises can further choose SaaS, VPC, on-premises, or air-gapped deployment.
Tabnine says code used for context is discarded after output generation and is not shared with third parties when using Tabnine models.
Tabnine describes ephemeral processing with no code storage on its servers after inference. Operational metrics and logs in self-hosted setups are described separately from code retention.
The reviewed privacy materials focus on ephemeral processing and training restrictions rather than one general human-review statement.
Foundational reference pages for teams adopting, deploying, and operating production AI with inspectable controls.