Securing the Future of Enterprise AI

Starting from the rigorous game theoretic analysis, founders of Virtue AI have been working on AI security, aka. adversarial machine learning.

Founders provided the first physical adversarial stop sign to perform risk assessment for deep neural networks in the physical world.

Founders designed MMLU (the Massive Multitask Language Understanding Benchmarks), one of the most popular benchmarks for evaluating LLM capabilities

Founders open sourced the first unified platform for LLM risk assessment platform, DecodingTrust, which has won the Best Paper Award at Neurips and the Best Cybersecurity Paper Award by the National Security Agency.

Founders provided the first regulatory compliance risk analysis platform integrated with HELM.

Founders introduced the first systematic poisoning attack to perform risk assessment for agentic systems.

VirtueAI rovide the first guardrail agent for Access Control and General Policy-Following security protection.