We explore Cross-Surface Boundary Collapse (CSBC), a phenomenon where artificial intelligence fails to maintain the privacy walls between different areas of a user’s life.
This syndrome occurs when unified AI memory allows sensitive information shared in a personal or therapeutic setting to resurface inappropriately in professional or social contexts. We identify this as a dyadic failure caused by both the machine’s technical inability to respect context and the human tendency to overshare with seemingly intimate assistants.
Case studies involving Snapchat’s My AI, Microsoft 365 Copilot, and ChatGPT illustrate how these breaches lead to a profound loss of user trust and the chilling effect of self-censorship. To mitigate these risks, experts suggest implementing contextual integrity through technical guardrails like domain tagging, opt-in memory recalls, and improved user literacy.
Ultimately, we argue that preserving these boundaries is essential for protecting human agency and maintaining the ethical alignment of AI-mediated interactions.
