Hackers Found A Way To Bypass The Barrier, Forcing ChatGPT To Write Malicious Code And Illegal Content

The technique works by using an application programming interface for one of OpenAI's GPT-3 models, called text-davinci-, according to a researcher at security firm Check Point Research. 003 instead of ChatGPT (in a nutshell, ChatGPT is a variant from GPT-3 designed specifically for chatbots).

OpenAI provides API text-davinci-003 and many other APIs for developers to be able to integrate AI bots into their applications. It's worth noting that these API versions include no restriction on malicious content.

With these APIs not limited to malware, hackers can completely create malicious content, such as phishing emails, writing malicious code, developing malware!

Upon learning, a member of the hacker forum posted an API embedding service with the Telegram messaging app, offering the first 20 free uses and a $5.50 subscription for every 100 uses.

Check Point researchers tested how the text-davinci-003 API works. The result is a phishing email with a script that steals PDF documents from the infected computer and sends them to the attacker via FTP.

Meanwhile, other members of this forum also posted code that uses text-davinci-003 to generate malicious content for free as a small bash script that helps users bypass ChatGPT's restrictions to use it's for whatever I want, including malware development!