Atuin is very clear about shell history being encrypted locally and not shared. I can’t find similar in relation the AI feature. Before using it I’d want to know if my prompts (and the history they contain) are retained, what LLM providers are they shared with, are they used to train future models, etc.
1 Like
Hey! Sorry for the delay in getting back to you, was wrapped up with 18.17. We will get a proper, comprehensive policy published on the site soon, but in the meantime:
- We retain nothing beyond summary usage data. This means your prompts are not stored, your shell history and other tool call output is not stored, etc. All that is stored is a summary of how many tokens were used, and for which model
- We switch provider regularly, however all providers we use follow zero data retention - no storage, no training, and are based in the US.
- Because of the above, there is no training on your data, as we do not store it
Additionally, we will be releasing a self hosted AI server in the coming weeks. This will allow you to use your own providers - cloud, local, or otherwise.