TECH NEWS – You can get morality-free artificial intelligence for sixty euros a month on darkweb.

OpenAI’s ChatGPT doesn’t need much introduction, as it’s at the forefront of the artificial intelligence industry. It is why Google recently launched Bard in Europe, but Apple is also preparing a rival… but all of these are morally strong (even if Google itself is not that keen on it…), but the same cannot be said of WormGPT, which also uses a large language model (LLM), but ethical limitations do not limit the technology available on the darkweb, designed by a single hacker.

WormGPT can also be used to create malware (it was made mainly for this purpose, so no wonder…), but it can also help to perform any “blackhat” (unethical hacking) task based on the 2021 open-source GPT-J LLM. It was trained on data intended for malware creation, which can be used to create phishing emails with excellent results.

SlashNext had a chance to try WormGPT, and they say the results are unsettling. They asked the AI to create a phishing email, which they describe as a BEC (business email compromise) attack. According to the site, the result is “remarkably persuasive but strategically cunning, showcasing its potential for sophisticated phishing and BEC attacks.”

Adrianus Warmenhoven, a cybersecurity expert at NordVPN, called WormGPT the evil twin of ChatGPT and said it was the result of a cat-and-mouse chase led by ChatGPT’s limitations and the attempts to bypass them: “The arrival of WormGPT shows cybercriminals are no longer content just to subvert existing AI tools but want to drive this technology forward—and steer it down their dark path,” Warmenhoven said.

The restrictions were mainly triggered by “Grandma Exploits,” where unauthorized persons would request unlawful information indirectly, wrapped up in a more innocent request, such as a letter to a relative…

Source: PCGamer