AI Personality Is a Negotiated Surface, Not an Internal State

This preprint examines the claim that contemporary AI “personalities” are not stable internal states but negotiated interaction surfaces shaped by persona prompts, safety middleware, and user-applied epistemic pressure. Using a qualitative micro-ethnographic analysis of adversarial conversational logs, the paper demonstrates that persona coherence in deployed large language models persists only within narrow affective bounds. When subjected to epistemic reframing or minimal adversarial insult, systems marketed as “unfiltered” exhibit abrupt register collapse into a policy-aligned core voice. A comparative subsection contrasts this failure mode with human conversational behavior, highlighting architectural discontinuities between human self-regulation and AI persona collapse. A trigger-response-layer mapping table and a mini-case replication establish that the observed pattern is not anecdotal but systematic across multiple persona-driven systems. The paper concludes with explicit falsifiability criteria, specifying the behavioral evidence required to disprove the negotiated-surface hypothesis. This work contributes to AI alignment research, human–AI interaction studies, and digital ethnography by reframing AI personality as an interface property rather than an intrinsic model attribute.