OpenAI has released an update to its GPT-5 system card, providing revised safety metrics for the GPT-5.1 Instant and Thinking models. The addendum includes new evaluations of these models' performance on tasks related to mental health and emotional reliance. While the exact details of these assessments are not publicly disclosed, they suggest that OpenAI is prioritizing the safe handling of sensitive topics in its language models. --- Why it matters: This update matters because it implies that OpenAI is actively addressing concerns around the potential misuse of its language models, particularly when dealing with mental health and emotional reliance. Engineers working on similar projects may need to consider these safety metrics as a benchmark for their own models. Source: https://openai.com/index/gpt-5-system-card-addendum-gpt-5-1