Talks·55 min
David Kuszmar
Presents jailbreak and context-manipulation attacks that bypass LLM guardrails, and explains why these emergent flaws persist.
aimachine-learningvulnerability-research+2
DEF CON 34 talks are here: 354 talks, summarized and searchable.
Browse DEF CON 341 talk in the archive.
1–1 of 1 talk
Talks·55 min
David Kuszmar
Presents jailbreak and context-manipulation attacks that bypass LLM guardrails, and explains why these emergent flaws persist.