Guardrails and risks: major chatbot safeguards appear to mitigate harm, suggesting current AI tools may not greatly enhance cybercriminal capabilities; however, poorly secured AI systems and agentic AI pose broader industry risks.

Researchers analyzed roughly 100 million posts from dark web and underground cybercrime forums since late 2022 to assess how cybercriminals are using artificial intelligence.

The study notes that cybercriminals are anxious about losing IT work to AI in mainstream software, a concern that could push some actors toward more criminal activity.

The findings are peer-reviewed and will be presented at the Workshop on the Economics of Information Security in Berkeley, California, this June.

The research has been peer-reviewed and will be presented at the Workshop on the Economics of Information Security in Berkeley, USA, in June 2026.

The study is peer-reviewed and will be presented at the Workshop on the Economics of Information Security in Berkeley, California, in June.

This material is a public release and the authors’ views are their own; further details are available via Mirage News and the linked full article.

Findings will be presented at the Workshop on the Economics of Information Security in June 2026; the lead researcher notes that the main danger lies in security risks from widespread AI adoption in industry, not solely in cybercriminal use of AI.

Emerging use cases involve automating complex tasks such as social engineering, phishing, and botnet management through AI-enabled automation frameworks, signaling incremental improvements rather than disruptive breakthroughs.

Overall, the research offers a nuanced, real-world view of AI in cybercrime: adoption is cautious and evolutionary, tightly tied to existing infrastructures, requiring robust security measures to prevent misuse.

Evidence suggests AI is aiding more advanced automation, particularly in social engineering and bot farming, but cybercrime remains heavily industrialized and driven by automated tools and pre-made assets, indicating evolution rather than revolution.