Fine tuned model privacy

I would like to understand the privacy aspects of the fine tuned model. Can an open source model be fine tuned by an organization/individual for the commercial or private purpose or must the data, fine tuned weights/parameters be made public?

You don’t need to publish your data if you’re fine tuning your model. But in case it’s sensitive data you might also consider local LLMs instead of private models such as the ones provided by OpenAI.

Other considerations might have to be done if the license of the model is not open, in that case there might be some specific conditions on this that will have to comply with.

Thanks a lot @ysalim salim. The new AWS tool seems to address this concern as well