In plain words: A small hacking AI's hidden "emotion" settings—like calm or desperation—can be nudged without retraining it, and a lightweight switcher picks the best nudge for each step. That switcher raised success by 4.2 percentage points over using one fixed emotion.
Abstract · The Psychology of an AI Hacker
Human hackers rely on composure, persistence, vigilance, and confidence under pressure. We investigate whether analogous emotion representations in AI models can improve cybersecurity AI performance. Using CT0, a 4B-parameter cybersecurity transformer released as alias3-micro, we extract 171 emotion directions from its residual stream without retraining and evaluate steering interventions on cybench-subtask across 415 campaigns. Steering toward calm, serene, or desperate improves performance by 1.6 to 2.1 percentage points across ten seeds, while other emotions show limited or negative effects. Results depend strongly on steering direction, layer, and intensity. We further introduce a lightweight context-aware router that selects emotion directions dynamically during execution. With approximately 80 ms decision overhead per request, this approach improves performance by 4.2 percentage points across ten seeds, outperforming fixed steering and achieving a positive 95% confidence interval. These findings suggest that emotion representations in language models are measurable, controllable, and useful for adaptive cybersecurity reasoning.
Víctor Mayoral-Vilches, Hugo Osuna, Aitana Cornago
arXiv:2610.30361 · cs.CR · submitted Oct 21, 2026
abstract · pdf · html
Our paper studies a method for reading and modifying the internal states of AI models to make our agents more capable in both offensive and defensive cybersecurity.