According to its report, “Inside CARLIY0S: How CyberCartel Used Multiple AI Models to Attack Latin American Banks,” Straiker, a cybersecurity company headquartered in Mountain View, California, that specializes in protecting AI applications and AI agents from cyberattacks, found that CyberCartel, also known as Tr4nsHack, allegedly used an AI-powered system called CARLIY0S to target more than 400 banks and financial institutions across Latin America.
How CARLIY0S Uses Multiple AI Agents to Plan Attacks
The system reportedly uses a “Cognitive Swarm” of four specialized Claude AI models that assess and debate potential targets before submitting their conclusions to a central arbiter. Once a decision is made, a commander coordinates the next steps. Other AI models, including Grok and Kimi, reportedly help automate target research, generate attack code, and analyze stolen data. The attackers allegedly bypassed safeguards by disguising malicious requests as authorized security testing.
Straiker’s findings show how AI can make cybercrime faster, cheaper, and easier to scale, letting small groups coordinate sophisticated attacks. The report warns that organizations remain vulnerable even if they don’t use AI themselves, because attackers can use AI-powered tools to target applications, APIs, and sensitive information.
