The Psicology of an AI Hacker

A paper submitted to arXiv on 21 October 2026 by Víctor Mayoral Vilches reports that steering a 4B-parameter cybersecurity transformer, CT0 (released as alias3-micro), toward calm, serene or desperate emotion directions improved performance by 1.6 to 2.1 percentage points across ten seeds on cybench-subtask, based on 171 emotion directions extracted from the model's residual stream without retraining over 415 campaigns. A lightweight context-aware router that selects emotion directions dynamically during execution added roughly 80 ms of decision overhead per request and improved performance by 4.2 percentage points across ten seeds, outperforming fixed steering with a positive 95% confidence interval. The results indicate emotion representations in language models are measurable and controllable for adaptive cybersecurity reasoning, though effects depended strongly on steering direction, layer and intensity.

Computer Science Cryptography and Security Submitted on 21 Oct 2026 Title:The Psychology of an AI Hacker View PDF https://arxiv.org/pdf/2610.30361 HTML experimental https://arxiv.org/html/2610.30361v1 Abstract: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. Submission history From: Víctor Mayoral Vilches view email https://arxiv.org/show-email/f7301a3b/2610.30361 v1 Wed, 21 Oct 2026 19:24:42 UTC 2,614 KB Additional Features References & Citations Loading... Bibliographic and Citation Tools Bibliographic Explorer What is the Explorer? https://info.arxiv.org/labs/showcase.html arxiv-bibliographic-explorer Connected Papers What is Connected Papers? https://www.connectedpapers.com/about Litmaps What is Litmaps? https://www.litmaps.co/ scite Smart Citations What are Smart Citations? https://www.scite.ai/ Code, Data and Media Associated with this Article alphaXiv What is alphaXiv? https://alphaxiv.org/ CatalyzeX Code Finder for Papers What is CatalyzeX? https://www.catalyzex.com DagsHub What is DagsHub? https://dagshub.com/ Gotit.pub What is GotitPub? http://gotit.pub/faq Hugging Face What is Huggingface? https://huggingface.co/huggingface ScienceCast What is ScienceCast? https://sciencecast.org/welcome Demos Recommenders and Search Tools Influence Flower What are Influence Flowers? https://influencemap.cmlab.dev/ CORE Recommender What is CORE? https://core.ac.uk/services/recommender arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs https://info.arxiv.org/labs/index.html .