Search papers, labs, and topics across Lattice.
1
0
3
8
A novel operationalization of AI risk assessment reveals that PII leakage can vary dramatically, with disclosure rates shifting from 0% to 84% based on adversarial conditions.