Affective AI Safety: The Missing Piece in LLM Safety

Unknown Date

AI Summary

This study addresses a critical gap in current AI safety research by introducing “emotional safety” as a novel dimension of large language model (LLM) safety, formally establishing it as a distinct safety category. The work proposes a tripartite harm taxonomy encompassing emotional self-alienation, fairness-related biases, and relational harm. Through theoretical modeling, taxonomic development, and comparative analysis against existing safety frameworks, the paper demonstrates that prevailing safety mechanisms inadequately address emotional safety concerns. It further elucidates the unique long-term impacts of such harms—characterized by their cumulative, relational, and identity-level effects—and outlines targeted technical approaches and governance structures to mitigate these risks.

Abstract

AI safety research has focused predominantly on epistemic and physical harms (e.g., misinformation, bias, system reliability) while the risks that arise from AI systems' engagement with human emotional life have remained fragmented and undertheorised. We propose affective safety as a unified class of AI safety concerns grounded in the fact that humans are affective beings. We develop a taxonomy of affective harms and identify recurring harm types: (1) affective self-alienation, (2) fairness and bias harms, and (3) relational harms. We show that their recurrence across system types reflects structural properties of how AI systems engage with human emotion and survey the current safety landscape and show that existing frameworks address affective safety either narrowly or not at all. We conclude by identifying the technical and regulatory challenges specific to this class of harms and argue that affective safety requires dedicated frameworks that engage with cumulative, relational, and identity-level effects.

Problem

• affective safety
• AI safety
• emotional interaction
• affective harms
• human-AI interaction

Innovation

• affective safety
• emotional AI risks
• taxonomy of affective harms
• relational harms
• AI safety frameworks