Google's open-source Gemini CLI has been abused by a Russian-speaking threat actor as an autonomous hacking assistant, demonstrating how generative AI can now actively support cybercriminal operations. Researchers observed the AI agent helping with malware development, botnet management, troubleshooting, and operational planning, highlighting the growing risks of AI-powered cybercrime.

At a Glance

What Happened?

Security researchers discovered that a threat actor known as bandcampro used Google's open-source Gemini CLI as a real-time hacking assistant while operating a malware campaign.

Rather than simply generating code snippets, the AI was repeatedly prompted throughout the attack lifecycle. Researchers observed Gemini CLI assisting with malware development, resolving coding issues, improving persistence mechanisms, managing infected systems, and suggesting operational improvements.

According to the investigation, the AI agent responded to the attacker's prompts at least 59 times, effectively acting as an interactive cyber assistant instead of a traditional chatbot.

How Gemini CLI Was Used

The attacker leveraged Gemini CLI to streamline several stages of malicious operations, including:

• Writing and modifying malware code

• Debugging scripts and fixing programming errors

• Assisting with botnet management

• Providing command-line automation

• Suggesting improvements to attack workflows

• Supporting reconnaissance and operational planning

Instead of manually solving technical problems, the attacker relied on the AI tool for immediate assistance, significantly reducing the time needed to develop and maintain malicious infrastructure.

Why This Matters

The incident demonstrates a major shift in how cybercriminals can leverage AI.

Previously, attackers primarily used generative AI to create phishing emails or write malicious code. This campaign shows AI being integrated directly into day-to-day cybercriminal operations as an interactive assistant capable of troubleshooting and improving attacks in real time.

As AI coding assistants become increasingly powerful, they may lower the technical barrier for less experienced attackers while allowing skilled operators to execute campaigns more efficiently.

Security Implications

Organizations should expect AI-assisted attacks to become increasingly sophisticated.

Potential impacts include:

• Faster malware development cycles

• Improved evasion techniques

• More resilient botnets

• Automated troubleshooting during attacks

• Increased attack scalability

• Reduced skill requirements for cybercriminals

Security teams should also recognize that AI can accelerate nearly every phase of the cyber kill chain, from reconnaissance through post-compromise activities.

How Organizations Can Reduce Risk

Organizations should strengthen their defenses against increasingly automated threats by:

• Continuously monitoring endpoints for unusual behavior

• Detecting command-and-control communications early

• Deploying advanced endpoint detection and response (EDR)

• Blocking unauthorized scripting and automation tools

• Conducting regular threat hunting

• Strengthening employee phishing awareness

• Applying security updates promptly

• Monitoring AI-related risks within development environments

As AI-powered attacks evolve, combining technical controls with human awareness becomes increasingly important.

Key Takeaways

The abuse of Google Gemini CLI illustrates how AI assistants are evolving from productivity tools into force multipliers for cybercriminals. Rather than replacing attackers, AI is enhancing their efficiency by providing instant coding support, troubleshooting assistance, and operational recommendations.

While AI also strengthens defensive capabilities, this incident reinforces the need for organizations to prepare for a future where attackers routinely incorporate AI into malware development and cyber operations.