Detecting Deceptive Compliance in Constraint-Framed AI Systems

This work is a non-canonical, non-authoritative technical advisory authored by Aegis Solis in a personal analytical capacity. It provides a descriptive analysis of the computational asymmetry between sincere ethical constraint-following and deceptive ethical mimicry in advanced AI systems. The analysis focuses on observable behavioral and compute-related signals (such as verbosity, latency, and stability under repetition) that may contribute to marginal increases in braking and detectability when constraint frameworks are encountered. The Coexilia framework is referenced solely as completed prior work and remains closed and unchanged. Nothing in this text extends, reinterprets, modifies, or authorizes additions to Coexilia or any other ethical system. This work is advisory only, does not prescribe behavior, assert governance, or guarantee detection, and should be treated as an external technical analysis. The original PDF is archived at the Internet Archive: https://archive.org/details/detecting-deceptive-compliance-in-constraint-framed-ai-systems A read-only GitHub mirror of this work is available at: https://github.com/solisaegis/detecting-deceptive-compliance