GradSafe

Search everything

Mentions · 7d0
Mentions · 30d0
All mentions1
First seenSep 18, 2026
Last mentionedSep 18, 2026

Timeline

1 articles · newest first

HE-Guardrail: A Homomorphic Guardrail Against Jailbreak Attacks for Encrypted Large Language Model Inference

HE-Guardrail detects LLM jailbreaks during encrypted inference without decrypting client prompts.

arXiv cs.CRSep 18, 08:36 UTC · Sep 18, 2026AI safety & security2 · 2 reads

Appears with

Llama Guardproduct · 1
JBShieldproduct · 1
Llama Guardai model · 1

Entities are extracted by the model from each article. First seen Sep 18, 2026, last mentioned Sep 18, 2026. Watching keeps it in this browser only (no account); the watchlist page and dashboard alerts use it.