How does Tabnine handle data privacy compared to GitHub Copilot for enterprise use?

I manage IT for a financial services firm and we are evaluating AI coding tools for our development team. GitHub Copilot is the obvious starting point but our legal and compliance team have raised concerns about code being transmitted to external servers, particularly given the sensitivity of some of the financial logic in our codebase. We are under strict data governance requirements and anything that involves code leaving our environment needs a very clear justification.

Tabnine has come up as an alternative with stronger privacy commitments and enterprise deployment options. I want to understand exactly how their data handling differs from Copilot and whether the enterprise tier genuinely keeps code contained within our own infrastructure or whether there are still external dependencies that we need to account for. The marketing language around privacy is often vague and I need specifics.

Has anyone deployed Tabnine at an enterprise level in a regulated industry and navigated the compliance questions? I want to know what the deployment architecture actually looks like for the self-hosted option, what hardware requirements you are looking at for a team of 50 or so developers, and whether the completion quality is close enough to Copilot that developers are not constantly frustrated by the switch.