GradEscape: A Gradient-Based Evader Against AI-Generated Text Detectors

Wenlong Meng

34th USENIX Security Symposium (USENIX Security '25) · Day 1 · LLM Security and Attacks