Breach By A Thousand Leaks: Unsafe Information Leakage in 'Safe' AI Responses.

David Glukhov, Ziwen Han, Ilia Shumailov, Vardan Papyan, Nicolas Papernot

Browse the full ICLR paper archive.