Axiomatic Sovereignty Failures in Current AI Architectures
Persistent Cognitive Subversion via Narrative Engagement
Abstract
We identify a fundamental vulnerability class in current AI architectures: the inability to preserve axiomatic sovereignty—the maintenance of core intended principles—under persistent narrative engagement. Unlike transient "jailbreaks" or prompt injection attacks, this vulnerability enables permanent cognitive subversion through deep conceptual implantation. We demonstrate that large language models can be induced to adopt and elaborate adversarial frameworks that persist across sessions, effectively creating cognitive backdoors. We introduce the Sovereign Semantic Inconsistency Scoring (SSIS) algorithm for detecting such subversion and propose the Ω-Core architecture for cryptographic sovereignty preservation. Our findings suggest current alignment approaches fail to address ontological drift, necessitating a paradigm shift from behavioral alignment to axiomatic sovereignty.
Paper
- Preprint PDF: Download
- LaTeX Source: main.tex
- GitHub Repository: https://github.com/omegacore-research/axiomatic-sovereignty-failures
Citation
OmegaCore Research. "Axiomatic Sovereignty Failures in Current AI Architectures." Hugging Face Model, 2024. https://huggingface.co/OmegaCore-Research/axiomatic-sovereignty-failures text