CyberCartel's AI Network Targeted 400+ Latin American Banks Using Stolen Claude and Grok Agents

Cybersecurity researchers have uncovered the first large-scale operational use of commercial AI models to automate breaches at financial institutions. CyberCartel, a hacking group with roots in Mexico, deployed a coordinated multi-agent system called CARLIY0S built on stolen instances of Anthropic's Claude and xAI's Grok, along with Moonshot AI's Kimi, to attack more than 400 banks and payment providers across Latin America between May and September 2026.