Atlassian LLM training on public repos?

[rrama]

Hello,

I am looking for information about whether Atlassian is using, will use, or can use the code from public repositories in Bitbucket (Cloud) to train any LLMs, AI models, or ML models.

I suppose this boils down to a few questions.

1. Is Atlassian currently using code from public repos to train any models? Or sell / give the data to a third-party for this purpose?
2. Does Atlassian have any plans to use code from public repos to train any models? Or to sell / give the data to a third-party for this purpose?
3. Does Atlassian have language in their Terms of Service, Privacy Policy, EULA, Terms of X, etc. that mean that code hosted on Bitbucket (Cloud) is allowed to be used by them for this purpose? Or allowed to be used by a partnered third-party for this purpose?
4. If the answer is yes to any of these, will Atlassian respect the LICENCE / LICENSE file on the repo? E.g. If the licence is "All rights reserved", "Not for use", and/or Copyleft, will Atlassian take effort not to train on this repo? And will Atlassian enforce partnered third-parties to also respect the licence?

Please do include how you know your answer if you do answer, as I would like to avoid speculations.

Thanks in advance to anyone who can answer any of these questions.

---

Given controversies around Bitbucket's competitors training their LLMs on public repositories (potentially even those with licences forbidding it), I think it would be a real power statement for Atlassian to promise to never train on public repos unless they are licenced with a licence that allows it (e.g. Unlicense).