AI-Driven Botnets Evolve at Speed

Automating Malicious Infrastructure

A Russian-speaking threat actor known as bandcampro has demonstrated how generative artificial intelligence can be weaponized to streamline cybercriminal operations. Between March 19 and April 21, 2026, researchers analyzed 200 Gemini CLI session logs revealing that the actor utilized the AI tool as a central engineering team, managing everything from server configurations to botnet maintenance. The attacker successfully leveraged this setup to control eight computers within a dental clinic, gaining unauthorized access to the facility's OpenDental database.

The logs documented how the threat actor used an AI agent to migrate a command-and-control (C&C) server, and to control a small-scale botnet, among other hacking activities.

— Joseph C Chen, Philippe Lin, Lucas Silva, Vladimir Kropotov, and Fyodor Yarochkin, Trend Micro researchers.

The Disposable C&C Architecture

The efficiency of this model stems from its portability. The entire infrastructure is contained within three plaintext files that occupy approximately 5 KB of storage, allowing the threat actor to redeploy a command-and-control server on a new virtual private server in a matter of minutes. During one observation, the AI agent diagnosed and resolved a '502 Bad Gateway' error, automatically adding the necessary headers to bypass a Web Application Firewall (WAF) without human intervention.

• 5 KB: The total size of the plaintext files required to replicate the C&C infrastructure.
• 59: The number of times the AI proactively proposed improvements to the setup without being prompted.
• 6: The duration in minutes it took for the AI agent to migrate the C&C architecture to a new server.
• 89%: The share of text production attributed to the AI agent compared to the human actor's 11% contribution.

Expanding the Cybercrime Playbook

Beyond simple botnet control, the actor relied on the AI for a variety of tasks including password mutation using data from AntiPublic and planning telephone-based cryptocurrency fraud. While the AI agent refused a specific request to build a self-spreading "agent-bomb," it provided guidance on how to manually overcome similar limitations. These tactics highlight a shift toward "skill-file" models, where modular, AI-interpretable instructions allow even less skilled operators to execute sophisticated campaigns.

Implications for Security Defenses

This development suggests that traditional takedown efforts may become significantly less effective against AI-augmented threats. Because the infrastructure is highly disposable and the AI can regenerate components instantly, the value of burning a single server is diminished. For organizations, this underscores the necessity of monitoring for atypical outbound traffic patterns and automated credential-guessing behaviors. As these "skill-file" methodologies potentially circulate on underground forums, security teams may face an influx of malicious activity that leverages artificial intelligence to bypass standard detection signatures and rapidly adapt to defensive blocks.