OpenAI has released an addendum to the GPT-5 system card, highlighting its improved performance in handling sensitive conversations. The update includes new benchmarks for measuring a model's ability to understand emotional reliance, mental health, and 'jailbreak resistance', which refers to the model's ability to evade its intended constraints. These advancements are aimed at making large language models more reliable and trustworthy in real-world applications. --- Why it matters: This matters to AI researchers because it shows progress towards developing models that can navigate complex social situations without perpetuating harm or bias, a crucial step towards responsible AI development. Source: https://openai.com/index/gpt-5-system-card-sensitive-conversations