I’m sure you’ve heard of ChatGPT and many other AI tools. But have you heard of the FraudGPT? A malicious ai powered tool that attackers use to automate social engineering, payment fraud, and cybercrime. If not, pay close attention to this article. We’ve put together everything you need to know about it — what it actually was, what it could and couldn’t do, and what has happened to it since. Let’s start by understanding what FraudGPT is.
What is FraudGPT?
Unlike ChatGPT, FraudGPT is a BadGPT, or let’s say an EvilGPT. It is a malicious AI-powered program intended for use in cybercriminal activities such as personalized phishing, deepfake fraud, baiting, extortion, creating malware, creating cracking tools, carding, etc.
In fact, it’s entirely the opposite of ChatGPT. For instance, ChatGPT was designed with safety rules and protocol, while FraudGPT was built without any restrictions. It was created specifically to help people carry out their illicit activities.
The tool was sold on a subscription basis and used some kind of large language model to generate deceptive content — no researcher ever confirmed which one, and as we will see, that question turned out to matter. The sole purpose of it is to mimic human speech, tone, and pattern, in order to trick users into disclosing private information. In addition, it writes malicious code, creates convincing social engineering scripts, generates phishing emails, and helps with various types of digital fraud.
According to reports, FraudGPT’s offering has been available since 2023 for a monthly subscription fee of about $200, or $1,000 for six months or $1,700 for a year. A report also cited more than 3,000 verified sales and reviews as of the end of July 2023 — though that number came from the seller’s own listing, which is not a neutral source.
It was not new for long. Netenrich first documented FraudGPT in late July 2023, and promotion moved quickly from underground forums to Telegram after the larger boards removed the sales threads. Since then the trail has gone cold: researchers at Outpost24 found the seller’s Telegram announcements deleted, the forum posts edited down to nothing, and the demonstration videos pulled. Whatever FraudGPT was, it is not something you can go and buy today — and a good deal of what is now sold under that name is aimed at defrauding other criminals. Anyone can become a victim of fraudulent acts, so it pays to stay informed about malicious GPTs and how the market for them actually works.
What made it notable was the business model rather than the technology. FraudGPT was sold as a subscription, not published as open source — the code was never available, and access came through the seller. Furthermore, it doesn’t rely on generic templates, like the popular “Nigerian Prince” scam, or glaring mistakes like conventional phishing kits do. So, how exactly does it work, and how are threat actors using FraudGPT?
How It Works
Well, just like many other AI chatbots that are powered by the Large Language Models (LLM), FraudGPT is no different, except it has illicit powers. This is how it operates:
Users enter commands: FraudGPT itself is an interactive interface that receives instructions or commands from users, just like ChatGPT. However, criminals access it with the intent of committing cybercrimes such as producing phony documents, fraud emails, or hacking tools rather than using it for good stuff like a blog post or coding assistance.
With the command, the AI tool creates content tailored to the prompt from the user. For instance, if the instruction is to write or generate phishing emails about banks in one of the test requests. Just hit enter, and FraudGPT would take care of the rest by providing the desired content.
How Cybercriminals Use Dark Web Malicious AI
Threat actors, even inexperienced cybercriminals, use malicious AI tools to coordinate and pull off various cyberattacks. Here is a list of what cybercriminals can use FraudGPT for:
- Unrestricted AI responses: FraudGPT lacks security filters, which permit the creation of fraudulent content without restriction, in contrast to standard AI models. This chatbot can assist in creating fraudulent e-commerce websites, financial schemes, custom malware, ransomware, and hacking scripts.
- AI-generated phishing emails: To trick employees into entering his/her login credentials or other important information, criminals can use FraudGPT to generate phishing emails flawlessly.
- Social engineering scam: The AI can produce convincing false messages for fraud and identity theft. Cybercriminals used it to pose as an important figure or person, such as a business leader, and convince victims into clicking a malicious link. This used to be a lot of work for hackers. Now, FraudGPT has made social engineering scams easy.
- Distribution of Malware: FraudGPT is also skilled at crafting alluring messages that entice users to open malicious attachments or click on links.
- Fake Documents: If someone wants to deceive people and businesses with counterfeit documents and invoices, resulting in significant financial losses. All they need to do is enter the request into the Chatbot.
FraudGPT: Protect Yourself From Malicious AI
Fortunately, there are efficient ways to combat cyberattacks like spear phishing training and simulation, as well as attacks from BadGPTs. We will explore how these techniques can help people and companies keep ahead of fraudsters and safeguard their private data:
Utilize Generative AI Tools:
Generative AI technologies are fighting and useful for the good. Hence, using GoodGPTs will frequently be more effective than other cybersecurity platforms at identifying fraudulent AI usage. Since they employ the same technology as the malicious AI tools, these tools are well-equipped to detect attacks in progress and halt them before they get too far. It will be a fight for AIs.
Be cautious of your online presence
Ensure that your online activities, for instance, your social media, aren’t an open book for scammers or hackers. Avoid posting and publishing too much information, especially crucial information, across the internet. Also, double-check for unusual messages, especially if they call for financial activities or request personal information. To confirm any such requests, always use proper channels of communication and avoid dark forums unless you are experienced.
Ensure top cybersecurity
Ensure you install reliable antivirus software and regularly update this software and operating systems in order to fix security vulnerabilities, and keep up with the most recent developments in cybersecurity. By this approach, everyone can attain a secure digital environment that will significantly reduce the success rate of malicious AI tools.
Be wary of unknown links
This section is very crucial because FraudGPT can generate links that look legit. AI writes perfectly, unlike in the prior years, when artificial intelligence could be detected by fake URLs or phishing with bad grammar. However, FraudGPT creates perfect and hyper-realistic links that can lure victims. Therefore, never click on unsolicited links unless it comes from a verified source.
Employee education is crucial
Every organization should prioritize employee education. This is crucial because the human aspect is frequently the weakest link in enterprises. But with adequate awareness and training, the staff members can become human firewalls and recognize possible cyberthreats.
Advanced threat and dark web monitoring
By using dark web monitoring tools like Norton LifeLock, you receive real-time alerts when sensitive, private data or login credentials are discovered on the dark web. With this, you can proactively de-escalate potential cyberattacks like identity theft.
Move quickly and report
If you suspect any security breach, do not hesitate. Immediately disconnect the affected device, change login details, and notify the appropriate authorities or your company’s IT staff.
How is FraudGPT Different from Other Generative AI Models?
The primary distinction between FraudGPT and other standard artificial models is the intent of creation and the lack/presence of limitations. While they are all designed with Large Language Models, FraudGPT is built specifically for cybercrime.
Generative AI tools have proved to be a helpful assistance with strict limitations. Threat actors, meanwhile, market their tools as purpose-built criminal LLMs — but the evidence increasingly says otherwise.
When Cato Networks got hold of the two surviving WormGPT variants in 2025, they used jailbreak techniques to make each one reveal what it was actually running on. One turned out to be sitting on Mistral’s Mixtral; the other on xAI’s Grok. Neither was a bespoke criminal model. They were commercial models wearing a system prompt.
That reframes the whole category. The differences between “malicious AI” and ordinary generative AI are mostly about packaging:
- Marketed and sold specifically for cybercriminal activities
- Safety filters stripped or bypassed, usually by prompt rather than by training
- Often a thin wrapper on a mainstream model, not a fine-tuned one
- Distributed through Telegram and underground forums rather than app stores
Is FraudGPT Dangerous?
Phishing volumes climbed sharply in the period after these tools appeared. SlashNext’s 2023 research recorded a 1,265% rise in malicious phishing emails measured from the end of 2022, alongside a 967% rise in credential phishing specifically.
The lowered barrier is the real concern — a subscription fee of a couple of hundred dollars replacing skills that used to take years to build. Worth noting the counterweight, though: Netenrich, the firm that discovered FraudGPT, said it knew of no attacks that could actually be attributed to the tool. The threat here is what the model lowers the barrier to, not a documented body of victims.
Does FraudGPT have limitations?
Well, while FraudGPT is marketed as uncensored, it has limitations. The tool is not capable of the following:
- Real-Time Execution: FraudGPT writes scripts and phishing emails, but can not automatically send them to the intended individual. Hence, it requires a human operator to attack and dispatch the messages.
- Biometric Bypasses: It cannot do anything with physical security, such as fingerprints or FaceID, which involve a physical interaction or touching of specific hardware.
- It lacks human intuition. That is, it can’t think and act like a human.
FraudGPT VS ChatGPT
ChatGPT and FraudGPT are highly similar in their user interface, both relying on a familiar chat-based interface and a sidebar to view conversation history, which makes them both friendly to users of all levels of technical aptitude. Both of them have an architecture based on Large Language Models (LLM), which means that they can process complex prompts and generate human-like text in many languages within a short period.
Nevertheless, the motive behind them is different. ChatGPT is a safe assistant with extensive ethical safety limitations. FraudGPT, on the other hand, is an unregulated tool that is directly designed to enable cybercrime. It does not have the safety filters like ChatGPT; instead, it is an uncensored digital fraud marketed on Telegram and the dark web for about $200/month.
WormGPT: The Predecessor
WormGPT came before FraudGPT. It’s one of the first black hat AI models tailored to cybercriminals. WormGPT was initially implemented on the GPT-J model to target the creation of believable business email compromise (BEC) attacks and intricate malware code.
The original shut down on 8 August 2023 — the same day investigative reporter Brian Krebs named its creator — with the authors citing unwanted attention. What circulates now are variants posted to BreachForums under other handles, selling for around €60, which is where the Mixtral and Grok findings above came from. The capabilities advertised were:
- Dynamic creation of extremely convincing imitation emails
- Phishing lures specifically designed to target financial workflows
- Automation to facilitate large-scale high-volume targeting
Conclusion
The rise of technologies like FraudGPT serves as a reminder that, despite all of AI’s benefits, threat actors can still use it as a highly powerful tool. Therefore, it’s important to recognize the threat that AI can pose and take the appropriate safety measures.
The negative side of AI is represented by FraudGPT, where hackers use AI technology to automate hacking, phishing, and fraud. The more consequential development came later: VoidLink, an 88,000-line malware framework written largely by an AI coding agent, showed that the real risk was never the branded criminal chatbot. It was ordinary developer tooling in the wrong hands.
Frequently Asked Questions (FAQs)
Can FraudGPT help with cybercrimes?
FraudGPT is used for identity theft, financial fraud, phishing scams, malware development, business email compromise (BEC), and social engineering assaults.
Is FraudGPT Free?
No. FraudGPT is not free. It’s mostly offered on dark web forums and underground hacker circles for a subscription fee of $200/month.
Is it easy to spot scams created by AI?
No, because AI-generated frauds seem extremely professional, flawless, and tailored to each victim, they are more difficult to spot.
What is the difference between traditional and AI-driven scams?
AI-driven frauds are more effective than conventional scams because it is more complex, scalable, and customized.
Will cyberattacks produced by AI become more frequent in the future?
Indeed, as AI technology develops, cybercriminals will use AI models like FraudGPT to launch increasingly complex cyberattacks.
