ChatGPT independently initiates a dialogue with users

Over the past few days, several ChatGPT users have reported strange experiences with the chatbot on Reddit. He started the conversation to ask about the user’s personal affairs. OpenAI says the error has been fixed, Futurism reports.

In one of the described cases, the chatbot started a conversation with the user to ask how his first week of training went. In another, he asked how the user was feeling and if he still had symptoms of the disease.

Screenshots of the chatbot’s strange conversations quickly gained attention on Reddit and X. Some mistakenly linked the incidents to the release of a new AI model, o1, with better reasoning abilities . However, OpenAI explained that the chatbot’s unexpected behavior is due to a memory function added earlier, which allows ChatGPT to save information about certain aspects of the users’ lives mentioned in the dialogue .

“This issue occurred when the model tried to respond to a message that was not sent properly and appeared to be empty. As a result, it either gave a generic response or used ChatGPT’s memory,” explained a company spokesperson, citing that the issue has been resolved.

It was previously reported that the Netherlands fined an AI company €30 million for collecting facial images.

We remind you that in the European Union, on August 1, the law on artificial intelligence began, which is called the first such regulatory document in the world.