Answers are grounded in this paper and cite their sources.
Generative artificial intelligence (GenAI) systems are increasingly used as interactive communication environments in which users submit messages and receive AI-generated responses containing personal information, confidential materials, original ideas, professional knowledge, and potentially sensitive content. Although contemporary messaging platforms have introduced mechanisms for restricting message forwarding, existing forwarding-control approaches primarily address conventional human-to-human communication, content dissemination, or platform-level forwarding limits. The emerging use of conversational AI introduces a distinct communication-control requirement: forwarding permissions must be applicable independently to both user-generated messages and AI-generated responses. This paper proposes Bidirectional Message-Level Forwarding Control for Generative Artificial Intelligence (BMF-GAI), a unified framework for selectively restricting the forwarding of individual messages exchanged between users and generative AI systems. The framework establishes a bidirectional permission model in which user messages and AI responses can independently carry forwarding restrictions, allowing users to protect their own submitted content and control the redistribution of AI-generated content within the capabilities of the platform. It further introduces message-specific protection, conversation-wide restrictions, asymmetric permission configurations, temporary restrictions, authorization-based forwarding, permission inheritance, restriction visibility, and policy enforcement mechanisms. Unlike conventional forwarding-rate limitations, BMF-GAI focuses on whether a particular message is authorized to be forwarded rather than how frequently a message has already been forwarded. The proposed architecture separates message ownership, forwarding authorization, policy representation, and enforcement, thereby supporting integration into different generative AI communication environments. A formal policy model, threat analysis, implementation architecture, and evaluation methodology are presented. The framework also distinguishes platform-enforced forwarding restrictions from absolute prevention of information reproduction, recognizing that external photography, manual transcription, and unauthorized extraction cannot be completely eliminated by application-level controls. This research establishes a conceptual and technical foundation for treating forwarding permissions as a first-class communication capability in generative AI systems.
No takes yet. Share an insight, caveat, or question.
Jincheng Zhang (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: