International Workshop on Designing and Measuring Security in Systems with AI (DeMeSSAI 2026)
MATRA: Modeling the Attack Surface of Agentic AI Systems -- OpenClaw Case Study
International Conference on Machine Learning (ICML)
Position: Safety Must Precede the Deployment of Open-Ended AI Agents
International Conference on Machine Learning (ICML)
Position: Trustworthy AI Suffers from Invariance Conflicts and Causality is The Solution
International Conference on Machine Learning (ICML)
Certified Circuits: Stability Guarantees for Mechanistic Circuits
Annual Meeting of the Association for Computational Linguistics (ACL)
ProxyPrompt: Securing System Prompts against Prompt Extraction Attacks
