ChatGPT will now compliment people more after its users rebelled.
OpenAI, the developers of the chatbot, announced a new update to the model that powers ChatGPT earlier this month. That model, known as GPT-5, was initially promoted by OpenAI as a major breakthrough that would move the system closer to being a broadly useful and perhaps even more human assistant.
But immediately many users found faults with the new system. And many of those complaints focused on its tone, which they said had become less friendly.
On users forums, many complained that they had become used to ChatGPT being social and positive with them, but that the new system’s personality was wrong.
In the days since, the company has revealed a host of both new updates and promised changes that are intended to restore that personality to the tool. That has included allowing people to set the system back to GPT-4o, the older model that was supposed to be made obsolete by the new update.
Now, OpenAI says that GPT-5 will compliment its users in an attempt to make it less “formal”.
“We’re making GPT-5 warmer and friendlier based on feedback that it felt too formal before. Changes are subtle, but ChatGPT should feel more approachable now,” the company said.
“You'll notice small, genuine touches like ‘Good question’ or ‘Great start,’ not flattery. Internal tests show no rise in sycophancy compared to the previous GPT-5 personality.”
OpenAI has had trouble with ChatGPT’s sycophancy in the past. The system has been criticised for being overly supportive of its users, including when they were asking about potentially dangerous behaviours, in a way that could endanger vulnerable people who are talking to the system.
That appeared to be one of the guiding principles behind the changes introduced with GPT-5. But in practice many users have complained that it has become too unfriendly and cold, along with a host of more technical complaints about the limitations of the system’s processing.
