Dear Sourcegraph Team,
As per Sourcegraph - Cody Enterprise Terms of Use, Sourcegraph collects prompts and responses:
For individual users accessing the Sourcegraph Cody extension via Sourcegraph.com, Sourcegraph also collects the following User Content to support and improve user experience, though none of it will be used to train any generally available models:
- User Prompts
- LLM Prompts
- Responses
I’m assuming “individual users” includes Cody Pro users. Is there a fixed retention duration for this data at the Pro tier? Similar to how OpenAI states that they persist prompts for 30 days.
1 Like
Hey @JeremyTCD you are correct, individual users does include Cody Pro users. We don’t currently have a guaranteed/specified retention timeline for these records, but our AI and legal teams are discussing that. I’ll be sure to add your question and thoughts to the conversation!
@jdorfman thanks for the clarifications.
Thought I’d add that I’d be happy to pay more for privacy - I’m currently trying Cursor for $20 a month. Cursor has a “privacy mode” where neither prompts nor responses are retained. Cody looks superior to Cursor, with unlimited Opus and Gpt4, so if Sourcegraph created a tier at that price point with similar privacy levels I’d be more than happy to pay for it.
I understand that due to the unlimited Opus and Gpt4, Sourcegraph needs to retain data to check for abuse and provide support. So a retention duration would make sense, like Supermaven’s 7 days or OpenAI’s 30 days.
You hit it on the head. Abuse has been a big issue, and the only way to battle it is to check prompts from bad actors to make sure there aren’t any false positives.
There is a thread in our company’s Slack going over this right now. I will follow up once I know more.
1 Like