GuardNet: Ensemble Strategies of Shallow Neural Networks for Robust Prompt Injection and Jailbreak Detection

In three linesGuardNet is a guardrail system using an ensemble of shallow neural networks (BiLSTMs, 47M parameters) to detect prompt injection and jailbreak attacks on LLMs. The approach prioritizes diversity of example coverage and threshold calibration over model scale. Performance: AUROC 0.747 on blind dataset (n=200), F1 0.92 on proprietary benchmark, ~50ms latency on CPU.

Summary generated by Claude — human-verified