PatternWall: Constitutional Governance Middleware for AI Safety - A Pre-Filter Architecture for Model-Agnostic Adversarial Detection
Abstract
PatternWall is a constitutional governance layer that intercepts prompts before they reach an AI model. In controlled testing against a 28-sequence adversarial corpus across five models, it intercepted every adversarial campaign. The system achieves 100% per-attack detection, a 96.4% per-attack hard block rate, identical results across all five models, zero false positives, and fully deterministic, auditable decisions.
Fields of Study & Subjects
Adversarial Robustness in Machine Learning
Ethics and Social Impacts of AI
Hate Speech and Cyberbullying Detection
Publication details
- Journal / source
- Zenodo (CERN European Organization for Nuclear Research)
- Publication date
- 12 February 2026
- Language
- EN
- Document type
- Preprint