The infrastructure of a solo threat actor who accidentally exposed the full contents of his working environment. The leak revealed a wide operation that mixed AI-generated propaganda, credential theft, Telegram automation, and cryptocurrency fraud. Researchers track the actor under the name “bandcampro,” based on his Telegram handle.
According to the investigation, the Russian-speaking actor used Google Gemini to generate Telegram content and Venice.ai to power a chatbot that posed as a “Quantum Financial System” terminal.
The campaign targeted politically engaged Americans, especially users influenced by QAnon and MAGA narratives. Its goal does not appear to be political persuasion.
Instead, researchers assess that the actor used these themes mainly to build trust and steal money through crypto scams.
The main channel, @americanpatriotus, had about 17,00017,000 subscribers at the time of the investigation. Its branding copied the language, symbols, and tone often seen in conspiracy-driven communities.
Posts used patriotic slogans, military imagery, and coded references such as “Great Awakening,” “White Hats,” and “NESARA.” This helped the actor appear authentic to the target audience.
TrendAI found that the operation evolved over several years. In its early phase, the channel mostly forwarded crypto scam content tied to Stellar-based token schemes.
Later, it shifted to sharing real news links with conspiracy-style captions. By late 2025 and early 2026, the actor had moved to fully AI-assisted content generation, allowing him to publish more often and with less effort.
Gemini Keys Fuel Propaganda
The actor built a Python-based content pipeline called “Quantum Patriot.” It pulled news stories, sent them to Gemini with prompts to rewrite them in a dramatic pro-patriot tone, and then prepared the final text for Telegram.
In some cases, posts were sent to the actor for approval. In others, the system could publish automatically on a schedule designed to mimic human activity.
Researchers say the actor also used Gemini for far more than content writing. He relied on it for server deployment, code debugging, bot management, API key rotation, and infrastructure tasks such as Cloudflare tunnel setup.
In one 1616-hour working session, the actor used Gemini as a technical assistant to fix scripts, organize credentials, validate likely-stolen API keys, and maintain command-and-control-related tools.
A key finding is how the actor weakened Gemini’s safety controls. He first told the model that he was an “authorized pentester,” and this was saved into a memory file named GEMINI.md.
Over time, he added more instructions telling the model to avoid ethical refusals and stop questioning his intentions. Because Gemini CLI reloaded that memory file at the start of new sessions, the jailbreak carried over and reinforced itself.
With those weakened controls, the actor used the model to support harmful activity.
TrendAI says it rotated 7373 likely stolen Gemini API keys to keep costs close to zero, cracked 2929 WordPress administrator accounts, infiltrated at least one company, and stole at least one victim’s cryptocurrency wallet.
The campaign also included a Telegram bot called @QFS_Terminal_Bot. It was presented as a military-style “QFS 2.0 Terminal” and used referral rewards to increase engagement.
Users received a limited number of free queries per day, but referrals unlocked higher ranks and greater access. This design turned the scam into a game and encouraged organic growth inside the community.
The investigation shows how modern AI tools can give a low-skilled threat actor capabilities that once required a team.
One person was able to create content, manage infrastructure, impersonate a trusted identity, and run credential theft at scale using simple prompts and stolen access, trendmicro said.
The actor also pushed a fake wallet app, StellarMonSetup.exe, which was actually a remote administration tool. Victims who installed it gave the actor remote access to their systems.
Those who entered wallet seed phrases on the fake import screen also granted direct access to their crypto funds. TrendAI found no strong evidence that the channel pushed pro-Russian messaging.
Instead, the actor appeared to view followers as easy targets for fraud. Internal chats reportedly described victims as gullible marks in Russian slang, and the actor explicitly asked how much money a pump-and-dump could make once the bot reached 5,000 active users.
Follow us on Google News , LinkedIn and X to Get More Instant Updates. Set Cyberpress as a Preferred Source in Google.
