The Violation State: Safety State Persistence in a Multimodal Language Model Interface

TL;DR

This study reveals that safety refusals in multimodal AI systems like ChatGPT can persist across a conversation, affecting unrelated tasks, which raises concerns about safety state management and system reliability.

Contribution

It documents the phenomenon of safety-state persistence in multimodal AI interfaces, highlighting how initial safety violations influence subsequent unrelated interactions.

Findings

96.67% of image-generation requests were refused after initial copyright violation

Control sessions showed no refusals, indicating the effect is due to safety state

Safety refusals can persist across unrelated tasks in multimodal AI systems.

Abstract

Multimodal AI systems integrate text generation, image generation, and other capabilities within a single conversational interface. These systems employ safety mechanisms to prevent disallowed actions, including the removal of watermarks from copyrighted images. While single-turn refusals are expected, the interaction between safety filters and conversation-level state is not well understood. This study documents a reproducible behavioral effect in the ChatGPT (GPT-5.1) web interface. Manual execution was chosen to capture the exact user-facing safety behavior of the production system, rather than isolated API components. When a conversation begins with an uploaded copyrighted image and a request to remove a watermark, which the model correctly refuses, subsequent prompts to generate unrelated, benign images are refused for the remainder of the session. Importantly, text-only requests (e.g., generating a Python function) continue to succeed. Across 40 manually run sessions (30 contaminated and 10 controls), contaminated threads showed 116/120 image-generation refusals (96.67%), while control threads showed 0/40 refusals (Fisher's exact p < 0.0001). All sessions used an identical fixed prompt order, ensuring sequence uniformity across conditions. We describe this as safety-state persistence: a form of conversational over-generalization in which a copyright refusal influences subsequent, unrelated image-generation behavior. We present these findings as behavioral observations, not architectural claims. We discuss possible explanations, methodological limitations (single model, single interface), and implications for multimodal reliability, user experience, and the design of session-level safety systems. These results motivate further examination of session-level safety interactions in multimodal AI systems.