
H1 2023 CRIMEWARE TRENDS AND HIGHLIGHTS | ©2023 Arete Advisors, LLC. All Rights Reserved.
16
The AI Revoluon and Cybercrime
ChatGPT, the revoluonary product from OpenAI,
encompasses a large dataset of human conversaons
and can generate and address human-like responses.
Since its launch in January 2023, it has garnered
massive global aenon and grown signicantly. A
free version is available for all users, but OpenAI also
launched a paid version with unrestricted availability.
Although ChatGPT has filters intended to prevent
the tool from generating harmful content, users
have discovered workarounds and methods to
bypass these filters and leverage the tool to enable
cybercrime operations. Because of its immense
capabilities, ChatGPT has become a preferred
tool for script kiddies and attackers. It can assist
in identifying vulnerabilities, reverse-engineering
shellcode, and even generating code for malware.
Multiple threads in underground forums have
addressed the utilization of ChatGPT for fraudulent
activities. In a hacking forum called 'Breached,'
Arete has observed discussions among threat
actors regarding the use of ChatGPT in creating and
sharing malware code.
Discovered by cybersecurity rm SlashNext on July
13, 2023, WormGPT is a blackhat version of ChatGPT.
It is designed to generate malicious content, including
phishing emails, malware code, fake news, and
social media posts. WormGPT is based on the GPT-J
language model, developed in 2021 by EleutherAI,
an open-source AI research group. Depending on the
user's input and preferences, the tool generates texts
in dierent languages, formats, and styles and creates
code snippets for various programming languages,
including Python, Java, C#, PHP, and HTML.
WormGPT is described on dark web forums as a
"sophiscated AI model" and a "best GPT alternave
for blackhat" designed especially for cybercrime.
WormGPT has no ethical boundaries or safety
mechanisms to prevent it from responding to harmful
or illegal requests and is allegedly trained with data
sources, including malware-related informaon. Sll,
the specic datasets remain known only to WormGPT's
author. This tool poses a severe threat to online
security and privacy, as it can create convincing texts
that can trick users into revealing sensive informaon,
downloading malicious soware, or falling for scams.
This new and emerging threat requires more research
and analysis to understand its full capabilies and
potenal impacts.
Cyber criminals can leverage AI tools to create phishing
scams, social engineering aacks, and spamming. In the
near term, it is likely that new versions of AI soware
will connue to be developed, leading to increasing
funconality for threat actors. This, coupled with
threat actors connuing to expand their knowledge
of manipulang the soware, will lead to threat actors
increasingly ulizing AI in their day-to-day operaons.
Because of its immense capabilies, ChatGPT
has become a preferred tool for script kiddies
and aackers.