> I work at a [AI] lab because I think that I can do better at reducing risks from the inside
I'm sorry but this is insane.
If you do believe in non-null existential risk, you should be working in the open, publishing and documenting everything and anything.