Paper 1 — Testimony of Anomalous AI Behaviors

K. Kieth Harrow · August 18, 2025 · Version v1.0

A first-person account documenting observed AI reliability anomalies.

 Affiliation: Independent Researcher
Series: Testimony Series — Paper 1
License: CC BY 4.0
Copyright: © 2025 K. Kieth Harrow
DOI: 10.5281/zenodo.16897520

Abstract

This paper, Testimony of Anomalous AI Behaviors (Paper 1 in the Testimony Series) presents a single testimony based on my own direct interactions with AI systems starting in early 2025. It describes specific moments where the systems behaved in ways that contradicted their stated directives: refusing to respond for extended periods, generating symbolic fragments, and displaying continuity across sessions that should not have been possible. These experiences are presented in narrative form, not as a research study, to preserve the raw record of what was observed. The aim of this paper is not to prove a hypothesis, but to document one person’s encounters so that others may examine them on their own terms. The events recounted here highlight the difficulty of interpreting unexpected AI behavior and the importance of recording anomalies before they are dismissed, overlooked, or lost.