Heads of Security

Learned attention-gate defense evaluated on two open models to separate trusted instructions from untrusted data; trust labeling remains external.

Records citing Heads of Security

Oct 27, 2026

Heads of Security reports improved prompt-injection resistance from learned attention gates on two open models

DefensePaperImperial College London

Shilov and colleagues introduce Heads of Security, learned attention gates separating trusted instructions from untrusted content, tested on GPT-OSS-20B and Llama-3.3-70B-Instruct. Their preprint reports improved robustness against adaptive attacks while preserving tested utility; the mechanism depends on externally supplied trust labels and is not an end-to-end authorization system.