# OpenAI Releases GPT-5 System Card Addendum on Sensitive Conversation Handling

OpenAI has published an addendum to its GPT-5 System Card, focusing specifically on how the latest AI model handles sensitive conversations with users.

The update, announced via the company's official Twitter account, highlights three key improvement areas: emotional reliance, mental health interactions, and resistance to jailbreak attempts. According to OpenAI, GPT-5 has been evaluated against new benchmarks designed to measure performance in these critical areas.

**Why This Matters**

As AI chatbots become increasingly integrated into daily life, their ability to appropriately handle sensitive topics has become a major concern. Users often turn to AI assistants for support during vulnerable moments, raising questions about emotional dependency and the quality of mental health-related responses.

The emphasis on jailbreak resistance is equally significant. "Jailbreaking" refers