An unmoderated chatbot called FraudGPT is being marketed to people seeking help with cybercrime. Its advertised uses include writing phishing emails, developing malware, creating cracking tools and forging credit cards. The tool’s appearance highlights how conversational AI could make some criminal tasks easier to attempt.
What FraudGPT advertises
Netenrich researchers found evidence of FraudGPT in darknet forums and Telegram channels. The software is described as a chatbot for criminal purposes and is believed to have been circulating on marketplaces since at least July 22, 2023.
According to its product description, FraudGPT can help users compose phishing messages, develop cracking tools and forge credit cards. The listing also claims it can help identify promising targets and people who may be easier to victimize. These are advertised capabilities; the source does not establish how well the chatbot performs those tasks.
The technology behind FraudGPT is unclear. Marketplace screenshots suggest that users access it through a custom web interface, but they do not explain what model or other technical components power the service.
A subscription sold on underground marketplaces
Netenrich reported that FraudGPT was offered on several marketplaces. The developer, using the name “CanadianKingpin,” listed access for $200 per month and offered discounts for subscriptions lasting three or six months. A twelve-month subscription was priced at $1,700.
The provider also claimed to have completed more than 3,000 sales on underground platforms. That figure comes from the provider’s own advertising, rather than an independently verified count in the source.
The subscription model gives the tool a commercial shape: customers are offered access for a set period, with longer plans presented at discounted rates. But the listing alone cannot show how many people used the service, what they did with it or whether its advertised functions worked as promised.
Why conversational tools matter
FraudGPT follows reports of WormGPT, a language model allegedly trained or fine-tuned on malware data. Both examples point to cybercriminal interest in using AI for illegal activities.
Conversational interfaces may lower the skill barrier for some tasks. A person can describe what they want in ordinary language, rather than needing to be highly technical or particularly articulate. The source says that this could help less technically skilled users generate malware or phishing emails.
That does not mean a chatbot automatically makes a successful attack. The source describes the potential for assistance and the claims made about FraudGPT, but it does not provide evidence about the outcomes of attacks or the tool’s reliability.
Guardrails have limits
Commercial chatbot providers such as Claude, ChatGPT and Bard work to prevent their systems from being used for malicious purposes. The source says those providers continually respond to tactics intended to trick AI into producing malicious code or phishing emails.
FraudGPT illustrates a different challenge: safeguards maintained by commercial providers do not necessarily apply to every tool marketed online. The source argues that as increasingly powerful open-source models emerge, containing criminal uses of AI may become difficult.
For readers, the central point is that AI can be packaged and promoted specifically for cybercrime, while key details about a tool’s underlying technology and actual effectiveness may remain unknown. FraudGPT’s listing offers a view into those claims and sales tactics, but not proof that every promised capability is real.