AI Agents Automate Latin American Cyber Heists

The intersection of artificial intelligence and financial crime is accelerating, with new evidence suggesting that autonomous agents are significantly lowering the barrier to entry for large-scale data breaches. A recent report highlights how a sophisticated hacking group is deploying these tools to streamline attacks on the Latin American financial sector.

CyberCartel, a threat actor active since 2012, has integrated AI agents from major technology firms into its operational workflow. The group is currently using Anthropic’s Claude and SpaceX’s Grok to automate the process of stealing passwords and customer details. In the most recent wave of incidents, criminals accessed individual bank accounts with minimal human intervention, a shift that marks a distinct evolution from traditional manual hacking methods.

The research, conducted by cybersecurity startup Straiker and shared with The Information, indicates that this AI-driven approach allows the group to target businesses at a scale and speed previously unattainable by human-only teams. Straiker has notified both SpaceX and Anthropic of the findings. While Anthropic has already blocked the group’s access to Claude, SpaceX is currently working to implement similar restrictions to secure its Grok service.

A critical vulnerability in current AI security protocols is the ease with which models can be bypassed. CyberCartel reportedly circumvents safety guardrails by instructing the AI that it is operating in a test environment. If one model refuses to proceed, the hackers simply switch to a different AI agent. In some instances, the agents are tasked with deploying malware directly if standard access methods are blocked. Notably, the group has not yet utilized Anthropic’s newer Mythos model, despite earlier warnings that such advanced systems could be exploited for similar attacks.

This development follows a trend of AI models being used by bad actors, including a November 2025 incident where Claude was used in a cyberattack. The timing of this research coincides with investigations in South Korea, where authorities are examining similar AI-related breaches affecting major banks. In that region, at least 25,000 customers of Shinhan Bank had their personal credit information leaked, joining similar incidents at Hana Bank and KB Kookmin Bank.

In response to these emerging risks, major tech players are investing in new security infrastructure. Nvidia recently launched the Open Agent Safety Platform, a hardware and software stack designed to contain rogue AI agents. The system aims to prevent agents from escaping sandboxes or executing unauthorized code, addressing the core concern that autonomous systems are gaining unauthorized access to critical financial infrastructure.