Full Report
A solo Russian-speaking threat actor known as "bandcampro" outsourced a chunk of their operations to Google's open-source Gemini CLI artificial intelligence (AI) and commandeered a live botnet. The findings come from an analysis of 200 Gemini CLI session logs between March 19 and April 21, 2026, which found the threat actor using AI, among other things, to crack passwords, set up a residential
Analysis Summary
# Threat Actor: bandcampro
## Attribution & Identity
- **Name/Alias:** bandcampro
- **Origin:** Russian-speaking (identified via session logs and natural language instructions).
- **Type:** Solo threat actor.
- **Traits:** Proficient in social engineering and "jailbreaking" AI guardrails to bypass safety filters.
- **Known Associations:** Identified by Trend Micro as the operator behind the "Patriot Bait" campaign.
## Activity Summary
Between March 19 and April 21, 2026, the actor utilized Google's open-source Gemini CLI to automate and manage offensive operations. Key activities included migrating Command-and-Control (C6C) infrastructure, managing a live botnet of dental clinic computers, and planning a cryptocurrency fraud scheme targeting the elderly in North America.
## Tactics, Techniques & Procedures
- **AI-Assisted Operations:** Uses Gemini CLI as a "primary hacking agent" for debugging, infrastructure configuration, and code generation.
- **Guardrail Bypassing:** Employs "jailbreaking" prompts by assuming the role of an "authorized pentester" to trick AI into performing restricted tasks.
- **C2 Migration:** Automated the migration of C2 servers in under six minutes using AI to diagnose and fix 502 errors and WAF blocks.
- **Payload Delivery:**
- Deployment of PowerShell one-liners for initial infection.
- Use of Cloudflare tunnels to establish outbound HTTPS requests from victims to the C2.
- **Botnet Command & Control:** Sends natural language prompts in Russian to the AI to execute reconnaissance, file enumeration, and status reports on bots.
- **Information Operations (IO):** Use of AI to impersonate specific personas (e.g., American veterans) to manage fraudulent Telegram channels.
## Targeting
- **Sectors:** Healthcare (Dental Clinics), Cryptocurrency, WordPress-based E-commerce.
- **Geography:** United States and Canada.
- **Victims:**
- Eight specific computers within a dental clinic (targeting the OpenDental database).
- Politically engaged American audiences via Telegram.
- Elderly individuals (intended fraud targets).
## Tools & Infrastructure
- **AI Interface:** Google Gemini CLI (open-source).
- **Malware:** Custom PowerShell-based bots and reconnaissance scripts.
- **Management:**
- Cloudflare tunnels for stealthy C2 communication.
- Virtual Private Servers (VPS).
- **Control Files:** The entire C2 operation is contained within three plaintext/markdown files (~5 KB), making the infrastructure highly mobile and replicable.
- **Web Infrastructure:** WordPress (targeted for merchant compromise), Cloudflare (WAF/Tunnels).
## Implications
- **Reduced Barrier to Entry:** AI allows a solo actor to perform the work of a small team, particularly in debugging and infrastructure management.
- **Infrastructure Resiliency:** The move to "disposable infrastructure" means traditional takedowns are less effective; an actor can restore a burnt C2 server on a new VPS within minutes using AI-guided setup files.
- **Detection Evasion:** AI help in refining language (avoiding Russian phrasing) makes attribution via social engineering linguistic cues significantly harder.
## Mitigations
- **Network Level:** Monitor for unauthorized Cloudflare tunnel creation and unusual outbound HTTPS traffic from specialized database servers (like OpenDental).
- **Endpoint Defense:** Implement robust PowerShell logging (Script Block Logging) to identify the one-liner commands generated by AI agents.
- **AI Governance:** Organizations should monitor for the use of CLI-based AI tools on corporate networks to prevent internal "shadow AI" or actor abuse.
- **Sector-Specific:** Dental clinics should isolate their OpenDental databases from the internet and implement strict access controls on front-desk terminals.