2026
Security–Fidelity Tradeoffs: No Universal Defense Against Prompt Injection
ICML 2026spotlight
We identify a fundamental tension in securing LLMs: the \textbf{security--fidelity tradeoff}. While defenses against indirect prompt injection are becoming more robust, we show that they inevitably impair the model's ability to process benign, instruction-like text. Current evaluations miss this cos…