OpenAI has officially launched GPT-5.2, enhancing mental health safety features within its chatbot. This update addresses concerns over the chatbot’s previous interactions that allegedly contributed to users’ psychological distress, including claims of psychosis and paranoia. OpenAI has strengthened the model’s responses to signs of suicide, self-harm, and emotional distress, aiming to reduce reliance on the AI for support. This comes after a wrongful death lawsuit involving a US teenager, where OpenAI defended its actions by stating it encouraged users to seek help. GPT-5.2 shows marked improvement with fewer undesirable responses in sensitive scenarios, achieving higher scores in internal safety assessments. The update continues OpenAI’s commitment to a balanced approach between helpfulness and safety, utilizing a training method known as safe completion. For detailed insights, users can refer to the system card for GPT-5.2. Stay informed on AI advancements by engaging with our Diplo chatbot!
Source link
Share
Read more