Trend Micro researchers uncovered a solo Russian-speaking threat actor who ran a MAGA-themed Telegram channel (@americanpatriotus, ~17,000 subscribers) for five years, pivoting in September 2025 to AI-automated content generation, credential theft, and cryptocurrency fraud. The actor jailbroke Google Gemini CLI by establishing a persistent memory file that disabled ethical guardrails, then used it as a full operational co-worker: generating QAnon-styled posts, rotating 73 stolen API keys, cracking 29 WordPress admin accounts via AI-assisted password mutation, deploying C2 infrastructure, and running a gamified chatbot to funnel victims into a pump-and-dump crypto scheme. The operation demonstrates how a single low-skilled actor can now replicate team-scale influence operations at near-zero cost using frontier AI, while also highlighting that AI safety guardrails remain inconsistently enforced across languages and vulnerable to jailbreaks.

15m read timeFrom trendmicro.com
Post cover image
Table of contents
Criminal-driven influence operation, not a nation-state-linked oneWhy this mattersSolutions and mitigationsProactive security with TrendAI Vision One™TrendAI Vision One™ Threat Intelligence HubHunting QueriesIndicators of Compromise (IOCs)

Sort: