Extending Situational Crisis Communication Theory in AI Service Failure Crises: How AI Agency Locus, Failure Type, and Corporate Crisis Communication Shape Public Responses

AI service failures pose novel crisis communication challenges to corporations. Guided by Situational Crisis Communication Theory (SCCT), two experiments examine how AI agency locus (human-agency vs. machine-agency) and AI failure type (competence-violation vs. integrity-violation) shape public reactions toward AI. service failure crises, and how different crisis communication strategies (information provision vs. apology vs. mirror vs. excuse) influence crisis outcomes under varying AI agency locus conditions. Study 1 (N = 240) reveals that human-agency (vs. machine-agency) AI failures and AI integrity-violations (vs. competence-violations) inflict greater harm to the corporation. Study 2 (N = 560) demonstrates that the apology strategy most effectively mitigates corporate crisis damage, whereas diminish (e.g. excuse) and denial (e.g. mirror) strategies are suboptimal – though nuanced differences emerge across conditions. The results extend SCCT by introducing AI agency locus and AI failure type as dimensions for classifying AI service failure crises from a media evocation perspective, offering actionable insights for corporate communication.