Toronto, June 20, 2024. AutoAlign, an AI safety company providing robust generative AI security solutions, announced that its Sidecar firewall is now available through NVIDIA NeMo Guardrails, part of the NVIDIA NeMo platform for building and deploying generative AI models. The company also released a white paper demonstrating its security effectiveness across major LLMs.
Sidecar technology operates alongside LLMs to provide dynamic security while evolving with rapidly advancing models. The integration with NVIDIA NeMo Guardrails allows enterprises to deploy secure custom chatbot applications with confidence.
“Legacy model security and safety solutions are failing to keep up with today’s risks, creating headaches for companies and major issues for customers.”
The white paper demonstrates Sidecar’s effectiveness on models including GPT-4 and Claude 3 Haiku. Testing showed GPT-4 blocks jailbreak attempts 100% of the time with Sidecar, up from 88.8%, and increases prompt injection mitigation from 14.3% to 100%. Claude’s jailbreak prevention improved from 98.3% to 100%, with prompt injection handling rising from 38.4% to 100%.
“Sidecar, with highly tailored Alignment Controls, proves that it increases LLM security against widespread attacks.”
“With AutoAlign’s Sidecar running on NVIDIA NeMo Guardrails, developers can build and run generative AI models with enhanced protection.”
