AI is becoming part of everyday criminal workflows – Help Net Security

A recent study analyzing conversations across cybercrime forums between January and July 2025 demonstrates a growing integration of Artificial Intelligence into criminal activities. The research, based on 2,264 messages from 1,661 contributors across 21 forums, identified four key themes: the repurposing of mainstream AI services like ChatGPT, DeepSeek, and Claude; the marketing of AI products tailored for criminal use (such as WormGPT and FraudGPT); the adaptation of AI models for specific operations like scamming and automation; and discussions surrounding the operational risks associated with using these tools.

While experimentation with mainstream AI is widespread, a significant portion of activity involves criminal AI brands, often functioning as wrappers around existing models. Experts suggest monitoring the prevalence of AI claims in underground sales listings as an indicator of broader adoption. The study highlights a faster integration of AI into fraud and social engineering operations compared to malware development, due to more accessible profit incentives and limited defensive AI protection for individuals.

Despite the enthusiasm, skepticism exists regarding the reliability of AI-generated code and concerns about operational security, with criminals cautioning against revealing sensitive information in prompts. Researchers recommend monitoring metrics like scam report volumes, phishing message interception rates, and coordinated campaign activity to detect AI-driven criminal activity. Overall, the findings suggest an early stage of AI integration into cybercrime, characterized by experimentation, marketing, and ongoing debate.

(Source：Plato Data Intelligence)