Many of you probably associate ChatGPT from OpenAI - a tool thanks to which we can discover the internet anew. By sending an appropriate query or issuing a specific and precise command of artificial intelligence, which is "on the other side of the screen", we can easily obtain information on various topics, e.g. programming, cooking, general information, you can also create abstract things, fictional stories, stories, etc. .

However, CyberArk tried something else. They noticed that the chatbot from OpenAI is able to create polymorphic malware, which is simply a "virus" that is able to change its code in real time to be undetectable for the object it attacks.

Chat has filters that prohibit creating such things, and researchers have been able to circumvent this simply by repeatedly insisting on chat to create malicious code. And it worked.

Źródło: instalki.pl