The clone story is the one I keep thinking about. Joe tested the impersonation risk, which is the one you can see coming.
In HR the quieter risk is retrieval, because a model with access to every document you have made will surface a grievance note to whoever asks the right question.
The clone story is the one I keep thinking about. Joe tested the impersonation risk, which is the one you can see coming.
In HR the quieter risk is retrieval, because a model with access to every document you have made will surface a grievance note to whoever asks the right question.
100% agree, but I do think it's good to test that! What happens if... in a safe space / sandbox?