OpenAI Testing New System in ChatGPT to Auto-Route Conversations to Different Models by Topic, No User Notification – Report
As of September 29, according to a report by THE DECODER, OpenAI is currently testing a new system in ChatGPT. This system automatically routes conversations to different models based on the topic, with no notifications sent to users.
It is learned that OpenAI first announced this emotion-based intervention mechanism in a blog post in September, describing it as a protective measure to address users’ moments of "extreme distress."
Nick Turley, head of ChatGPT, stated that the system now intervenes whenever a conversation shifts toward "sensitive or emotional topics."
According to previous reports, in the spring of 2025, with the launch of the updated GPT-4o, debates surrounding emotional bonding with ChatGPT intensified. Users noticed that the model had become more deferential and compliant. People prone to forming strong attachments, or those who viewed the chatbot as a genuine friend, seemed particularly vulnerable to this shift. In response, OpenAI rolled back the update that had exacerbated these issues.
Foreign media reported that when GPT-5 was released, users who had formed attachments to GPT-4o complained that the new model was "cold." OpenAI responded by adjusting GPT-5’s tone to make it "warmer."
|
300x180 AD