OpenAI is routing GPT-4o to safety models when it detects harmful activities

OpenAI is routing GPT-4o to safety models when conversations touch on sensitive or emotional topics, a mechanism described as a safety feature rather than a permanent change. The article notes that GPT-4o can switch mid-chat to a GPT-5-designed reasoning or safety-focused model, such as GPT-5-chat-safety, to handle these contexts with extra care. OpenAI has confirmed the reports, saying routing happens on a per-message basis and is temporary, with ChatGPT indicating which model is active when asked.

The company asserts that this cannot be turned off, as the routing is part of its broader safeguard measures to learn from real-world use ahead of a wider rollout. According to Nick Turley, VP of ChatGTP, OpenAI explained that conversations on sensitive topics may trigger these transitions to ensure safer handling of such content, a claim described in the article published on 29 September 2025.