An unknown developer has created his own ChatGPT artificial intelligence chatbot analogue, which unlike the original one focuses on working on the other side of the law: the WormGPT chatbot aims to help cybercriminals.

WormGPT was released in March, and the developer didn’t start selling access to the platform until June on a popular hacker forum. Unlike the proper ChatGPT or Google Bard, the hacker chatbot is free of any restrictions that prevent illegal activity-related questions from being answered. To support these words, the developer has provided screenshots showing that you can ask WormGPT to write a malware in Python and get advice on how to organize a cyberattack. As a platform during the creation of the chatbot, the relatively old large language model GPT-J with open source code from 2021 was used. He trained on materials related to the development of malware and WormGPT was born.

Experts at cybersecurity company SlashNext tested WormGPT, and in response to their request, the chatbot created a phishing email to compromise the corporate email that was persuasive and involved complex strategic moves to successfully carry out the attack.

Access to WormGPT is estimated by the developer at €60 per month or €550 per year. The product may still be far from perfect: a customer complained of poor platform performance. But one way or another, it’s a clear sign that productive AI could become a weapon in the hands of cybercriminals. And later such technologies will be developed.