Research Insights

From academic research to production-grade AI safety.

Latest Research Papers

When Do Universal Image Jailbreaks Transfer Between Vision-Language Models?

Data Distillation Can Be Like Vodka: Distilling More Times For Better Quality

Identifying Spurious Biases Early in Training through the Lens of Simplicity Bias

SmallToLarge (S2L): Scalable Data Selection for Fine-tuning Large Language Models by Summarizing Training Trajectories of Small Models

DecodingTrust: A Comprehensive Assessment of Trustworthiness in GPT Models

TextGuard: Provable Defense against Backdoor Attacks on Text Classification

Can Pruning Improve Certified Robustness of Neural Networks?

Shake to Leak: Amplifying the Generative Privacy Risk through Fine-tuning

Improving Privacy-Preserving Vertical Federated Learning by Efficient Communication with ADMM.

Ring-A-Bell! How Reliable are Concept Removal Methods For Diffusion Models?

Blogs

See all Blogs ->

Product Announcement

Introducing Agent ForgingGround with Built-In Red-Teaming Agents

Enterprise-Scale Validation for Secure AI Agent Deployment. Enterprises are rapidly moving AI agents from experimentation into real business operations. These agents can call tools, access sensitive data, and execute critical actions across enterprise systems.

Business

Virtue AI Named as a Representative Vendor in the Gartner® Market Guide for Guardian Agents

Virtue AI has been recognized as a Representative Vendor in the 2026 Gartner® Market Guide for Guardian Agents. Thank you to our customers, partners, and team for making this possible!

Agent

February 9, 2026

Openclaw Is A High-Value Target. Here’s How To Secure It.

OpenClaw is exploding right now. In the past week, OpenClaw has gone from “interesting agent demo” to mainstream viral tool, praised for actually executing tasks across real apps (chat, email, workflows) rather than just chatting. But what happens when users grant this agent unlimited access?

Strengthen Your AI Posture Today

Virtue AI brings control, governance, and resilience to enterprise AI.