LLM Guard by Protect AI is a suite of tools designed to protect LLM applications by detecting, redacting, and sanitizing LLM prompts and responses for real-time safety, security, and compliance. It addresses significant security risks such as malicious prompt injections and data leakage in Retrieval-Augmented Generation (RAG). LLM Guard also tackles governance and policy concerns related to the sharing of sensitive data in LLM outputs. It includes built-in sanitization, detection of harmful language, prevention of data leakage, and resistance against prompt injection attacks. LLM Guard ensures that user interactions with LLMs remain safe and secure. It can be deployed on any LLM, including GPT, Llama, Mistral, Falcon, and in any LLM framework such as Azure OpenAI, Bedrock, Langchain, and more. LLM Guard is easy to deploy, integrating within minutes as a library or API, and includes extensive documentation with use-cases and playbooks. It features advanced input and output scanners that anonymize PII, redact secrets, and counteract threats such as prompt injections and jailbreaks. LLM Guard is optimized for cost-effective CPU inference, reducing inference expenses significantly compared to GPU. It is a permissively licensed open-source project, promoting rapid innovation and transparency, with a commercial version offering expanded features within the Protect AI platform coming soon.
