I am kinda confused regarding the privacy policy. in this post the support person says that in pro version cody wont train on private code, but does not mention about LLM providers (anthropic, openai).
So my questions are:
1a: Will Anthropic/OAI train on code/prompts supplied to them via cody?
1b: Does source graph train on code/prompts supplied to them via cody?
2: Do the not training gurantees apply for free users as well or just pro and enterprise users?
Hey @gj-raza
Indeed, Sourcegraph does not train on your prompts or code on since we don’t host them. We have a fine-tuned model for autocompletion trained on publicly available code.
Third-party LLM providers might train on your code if you are a Free/Pro user, because the prompts are redirected to them since we do not host LLMs by ourselves.
As an Enterprise tier user, you have different conditions as multi-tenant self-hosting or cloud-hosting options and therefore a different policy with the third-party LLM providers.
Our telemetry system is solely for gathering metrics about usability of our features in the products we provide. They stay in-house for evaluation purposes only.
I hope that helps clarify your questions.