Sensitive Data Leakage in Enterprise AI Workflows

[J1ank3R]

I’ve been testing A.I.G in an enterprise setup, and one thing I’m not sure how to evaluate is sensitive data leakage.
When an AI app has access to internal data, some of that information has different sensitivity levels, and not everything should show up in the model’s output.

It would be helpful if A.I.G could do some basic checks here — for example, warning when a response might include data that’s considered too sensitive for that workflow.

Just wondering if this kind of sensitive-data leakage assessment is something the maintainers are thinking about.

[boy-hack]

How to determine if it's a sensitive file could be an issue, and our team will discuss it.