Overly flattering: ChatGPT version withdrawn

A new version of the AI chatbot ChatGPT has been withdrawn for an unusual reason: It was too nice to users.

The variant of the GPT-4o model released just a few days ago had been overly flattering and had also been described as submissive, the developer company OpenAI explained the move.

Over the past few days, there have been an increasing number of examples online of how the ChatGPT version also euphorically supported definitely bad ideas – such as hugging a cactus. Users were also initially showered with praise even for simple requests.

Such behavior of the software could be unpleasant and disturbing for users, OpenAI justified its intervention. The “default personality” of ChatGPT influences how people interact with it and whether they trust the program.

Users like encouragement

One cause of the problem seems to be that the version was designed to please users – and according to the data, it was always well received by them.

OpenAI is planning more options for adapting the software’s behavior in the future – for example, with a choice between different personalities. It is already possible to control the tone of voice with instructions.

ChatGPT is the chatbot that triggered the euphoria surrounding artificial intelligence more than two years ago. The software now has around 500 million active users every week.

dpa