01▲HE-Guardrail: A Homomorphic Guardrail Against Jailbreak Attacks for Encrypted Large Language Model Inference arxiv.org Can jailbreak filtering work when the prompt stays encrypted? HE-Guardrail says yes, by running guardrails over ciphertext.arxivfhehomomorphic-encryptionjailbreakllmmachine-learningprivacysecurity0 pts/jonasdunn/7 days ago/1 comment