Jailbroken AI Russian cyber campaign was deployed in an advanced operation targeting U.S. politicians and politically engaged individuals, according to new findings from TrendAI. The campaign reportedly included phishing, fraud and AI-generated influence operations, where large language models were used to produce content, automate workflows and scale cyberthreats on a much larger scale.
New information from TrendAI indicates that jailbroken large language models (LLMs) were leveraged to construct sophisticated cyberthreats aimed at American politicians and politically active people. The operation is said to have combined AI-generated influence messaging, credential theft, phishing and large-scale cryptocurrency fraud.
According to the research, the campaign—identified as brandcampro—was run by a single Russian-speaking threat actor. TrendAI’s investigation suggests the operation relied on a jailbroken version of Google Gemini and other AI-based tools to automate content creation, build phishing infrastructure, harvest login credentials and execute crypto-related scams.
AI used to scale the cyber campaign
TrendAI’s analysis shows AI was not only used to generate content but also to streamline and scale the entire cybercriminal infrastructure behind the operation.
The campaign reportedly used AI to:
• produce convincing political content and influence messages
• automate phishing campaigns and create realistic fake login pages
• process stolen API keys and automate backend workflows
• carry out larger-scale cryptocurrency fraud with reduced operational costs
The report states the campaign built an audience of about 17,000 followers on Telegram, while AI-driven automation tools and stolen API keys significantly cut the cost of running attacks.
A new form of AI-driven cybercrime
TrendAI argues that brandcampro demonstrates how modern AI systems can enable a new generation of scalable cyberthreats that combine influence operations, automation and financial crime within the same campaign.
The case of this jailbroken AI Russian cyber campaign illustrates how AI can scale cybercrime in ways that previously required far greater resources.
“This cyber campaign shows how modern AI systems enable a new, scalable and cost-effective form of cybercrime that combines influence operations, automation and financial fraud,” says Martin Fribrock, Country Manager Sweden, Finland and Baltics at TrendAI.
Jailbroken AI models introduce new risks
The incident highlights growing concerns about so-called jailbroken AI models—where safety constraints in large language models are bypassed to produce content or tools that would normally be blocked.
When such models fall into the hands of threat actors, AI can do more than generate text: it can help build phishing pages, automate attack chains, generate fake identities and fuel sophisticated fraud campaigns at a fraction of the previous cost.
Security experts warn that jailbroken language models can enable a novel class of cyberthreats in which disinformation, phishing, financial fraud and digital influence are combined into unified operations.
A shifting threat landscape for AI security
Experts caution that jailbroken language models could let attackers automate phishing, fraud, disinformation and digital influence at much greater scale than before.
By using AI to create content, build convincing fake login pages and drive automated workflows, cybercriminals can lower costs while increasing the sophistication of attacks and making them harder to detect.
For companies, government agencies and political organizations, this trend makes AI security an increasingly vital component of cybersecurity efforts. Defending against these threats requires not only protecting data but also understanding how generative AI can be used offensively by adversaries.
The development also underscores that AI security is no longer solely a technical issue for IT departments. It is becoming a strategic concern for leadership, risk management and national security.
Read TrendAI’s research on brandcampro here: