Addendum to GPT-5 System Card: Sensitive conversations

OpenAI has released an addendum to the GPT-5 System Card detailing significant improvements in handling sensitive conversations, establishing new benchmarks for emotional reliance, mental health support, and jailbreak resistance across their artificial intelligence platform. The document represents the company's ongoing commitment to addressing the complex ethical challenges that arise when AI systems engage with human users on emotionally charged topics. By implementing these enhanced protocols, OpenAI aims to ensure that GPT-5 can appropriately respond to users' emotional needs without fostering unhealthy dependencies while maintaining robust safety measures against malicious manipulation. The addendum specifically outlines new performance metrics that the AI must meet when discussing mental health topics, demonstrating increased sensitivity to vulnerable users while avoiding the provision of potentially harmful advice. This development comes as part of OpenAI's broader transparency initiative, providing stakeholders with detailed insights into both the capabilities and limitations of their advanced language model.