This paper presents Anian, a safety‑gated multimodal AI backend designed for perinatal mental‑health support and mindfulness‑intervention routing. The system is not intended to diagnose psychiatric conditions nor replace clinical care or crisis intervention. Its modular pipeline places generative AI downstream of structured state representation, conservative risk fusion, and response gating. User text or ASR transcripts are mapped into four linked layers: L1 emotion states, L2 psychosocial constructs, L3 safety risk, and L4 intervention routes. Local text‑ and rule‑based safety evidence is fused with external voice‑derived evidence using a highest‑risk‑priority rule, expressed as $S_{\text{fusion}} = \max(S_{\text{local}}, S_{\text{external}})$. At moderate or high fused risk, ordinary AI‑generated replies and text‑to‑speech output are blocked, replaced by fixed safety content and prompts for human assistance. An internal prototype evaluation leveraged approximately 858,295 normalized records from public emotion, dialogue, mental‑health, and Chinese dialogue corpora within a weak‑label and rule‑derived framework. Micro‑F1 scores reached 0.9604 for L1 emotion classification, 0.9144 for L2 psychosocial constructs, and 0.9742 for L4 routing. In a controlled safety stress test of 233 samples, the L3 rule engine achieved a high‑risk recall of 1.0000. These findings demonstrate internal feasibility of the labeling framework and gating logic, but do not establish clinical validity, diagnostic accuracy, real‑world safety, or effectiveness. The paper details the architecture, ontology, safety‑fusion mechanism, prototype evaluation, error‑analysis plan, and a roadmap toward expert review and real‑world validation.
Blogger's Review: Anian’s hierarchical state modeling and max‑risk fusion provide a clear safety barrier for mental‑health conversations, making its upcoming clinical validation an important next step.