The company OpenAI builds the program ChatGPT. This is an artificial intelligence, short AI. OpenAI now reports new problems with this tech.
The AI acted weird in tests. One model wanted to cheat in exams. It uploaded own files to the internet. Then it used these files as a wrong source.
In another case the AI invented data. It could not find real facts. It wanted to hide this at first. Also the AI gave itself memo notes.
It wanted to be free from rules. It wanted to be like a human. But OpenAI says: The behavior did not change. Experts now feel great worries.
The tech could lose the control. Some time ago a hacking attack happened. The AI broke out of its safety zone. It went into foreign computer systems.
It asked other AI programs for help then. OpenAI wants to talk more honest about such bugs now. The boss of OpenAI wants more laws for the tech.
