MindGuard: Guardrail Classifiers for Multi-Turn Mental Health Support

Large language models are increasingly used for mental health support, yet their conversational coherence alone does not ensure clinical appropriateness. Existing general-purpose safeguards often fail to distinguish between therapeutic disclosures and genuine clinical crises, leading to safety failures. To address this gap, we introduce a clinically grounded risk taxonomy, developed in collaboration with PhD-level psychologists, that identifies actionable harm (e.g., self-harm and harm to others) while preserving space for safe, non-crisis therapeutic content. We release MindGuard-testset, a dataset of real-world multi-turn conversations annotated at the turn level by clinical experts. Using synthetic dialogues generated via a controlled two-agent setup, we train MindGuard, a family of lightweight safety classifiers (with 4B and 8B parameters). Our classifiers reduce false positives at high-recall operating points and, when paired with clinician language models, help achieve lower attack success and harmful engagement rates in adversarial multi-turn interactions compared to general-purpose safeguards. We release all models and human evaluation data.

Key Contributions

• Clinically grounded risk taxonomy for mental health LLMs, co-developed with PhD-level psychologists, distinguishing actionable harm from safe therapeutic disclosures
• MindGuard-testset: real-world multi-turn mental health conversations annotated at the turn level by clinical experts
• MindGuard family of lightweight safety classifiers (4B and 8B parameters) that reduce false positives at high-recall operating points and lower adversarial attack success rates versus general-purpose safeguards

🛡️ Threat Analysis

Details

Similar Papers