Show HN: RAXE Open Source – LLM Prompt Threat Detection (EmbeddingGemma L2) (github.com)

RAXE has launched its Open Beta v0.0.1 Community Edition, a free tool designed to enhance the security of AI systems by addressing vulnerabilities like prompt injections and data leaks, which have surged alongside the widespread deployment of large language models (LLMs). This beta release allows researchers, developers, and security teams to scan prompts locally for potential threats, ensuring that no data leaves their servers. RAXE’s architecture incorporates 460+ detection rules across seven threat families, combining regex-based and machine learning classifiers to provide dual-layer protection with less than 10ms latency. Significantly, RAXE emphasizes transparency and community-driven defense, allowing users to audit detection logic and understand the rationale behind flagged prompts. This mirrors practices in traditional cybersecurity, promoting shared intelligence within the AI community to improve safety measures collectively. With its open design, RAXE aims not only to block threats but also to foster a research-first environment where users can contribute detection rules and enhance the tool's accuracy. This alignment of the AI and security sectors highlights a crucial step towards robust safeguards in an increasingly digital landscape.