Anthropic
Anthropic and Claude are major enterprise-AI alternatives for a buyer who needs frontier-model capability under commercial privacy terms.
Privacy model
Anthropic says commercial customers are controllers and Anthropic is a processor for their commercial data. It also says it does not train generative models on commercial data.
Enterprise retention can be configured, but enterprise chat is stateful by design. API inputs and outputs are normally deleted within 30 days; flagged conversations and classifier records can have longer retention periods. Zero Data Retention requires an eligible agreement.
Lesson for the venture
No-training language and a DPA are valuable, but a hosted Claude request remains live plaintext in the provider’s model environment. Show that tradeoff honestly as a contractual external route, with its configured retention and tool boundaries. Private AI trust boundaries explains the distinction.