OMNIGUARD: An Efficient Approach for AI Safety Moderation Across Languages and Modalities

TL;DR

Omniguard is a novel, efficient method for detecting harmful prompts across multiple languages and modalities, significantly improving accuracy and robustness over existing approaches.

Contribution

It introduces a cross-language and cross-modality detection approach that leverages aligned internal representations for improved harmful prompt classification.

Findings

11.57% accuracy improvement in multilingual detection

20.44% accuracy increase for image prompts

State-of-the-art performance for audio prompts

Abstract

The emerging capabilities of large language models (LLMs) have sparked concerns about their immediate potential for harmful misuse. The core approach to mitigate these concerns is the detection of harmful queries to the model. Current detection approaches are fallible, and are particularly susceptible to attacks that exploit mismatched generalization of model capabilities (e.g., prompts in low-resource languages or prompts provided in non-text modalities such as image and audio). To tackle this challenge, we propose Omniguard, an approach for detecting harmful prompts across languages and modalities. Our approach (i) identifies internal representations of an LLM/MLLM that are aligned across languages or modalities and then (ii) uses them to build a language-agnostic or modality-agnostic classifier for detecting harmful prompts. Omniguard improves harmful prompt classification accuracy by 11.57\% over the strongest baseline in a multilingual setting, by 20.44\% for image-based prompts, and sets a new SOTA for audio-based prompts. By repurposing embeddings computed during generation, Omniguard is also very efficient ($\approx\!120 \times$ faster than the next fastest baseline). Code and data are available at: https://github.com/vsahil/OmniGuard.