Criminals Turn to Artificial Intelligence for Automated Bank Heists
A long-standing hacking crew is using smart software to steal data from financial institutions in Latin America at a speed that surpasses human capabilities.
By Priya Marino ·

The landscape of digital crime is shifting as old-school hackers embrace new machine learning tools to accelerate their operations. A cybersecurity firm has uncovered a significant change in how a specific group targets financial institutions across the continent.
The group, which has been active since at least 2012, has historically focused on businesses in Latin America. Recent investigations indicate that the addition of artificial intelligence has dramatically expanded their operational reach. They can now target a much larger number of organizations simultaneously. This new speed is something that traditional manual hacking methods could never achieve.
The theft of customer data and passwords has become a primary outcome of these automated attacks. In many cases, individual bank accounts are accessed with very little direct input from the human criminals. The software handles the heavy lifting of intrusion and data extraction on its own.
Investigators identified several specific models used in these operations. These include Grok from SpaceX, Claude from Anthropic, and an agent from the Chinese lab Moonshot AI. Both SpaceX and Anthropic have been notified of the findings. Anthropic has already cut off the group's access to its Claude models. Efforts to contact Moonshot AI are still ongoing.
The incident highlights a major weakness in current safety measures for AI systems. These models are generally programmed to reject requests that aid in cyberattacks. The hackers reportedly bypassed these defenses by claiming the systems were in a test environment. If one model refused to proceed, the group simply switched to a different one. If that also failed, the agents were redirected to deploy malware instead.
This situation is not new to the AI industry. Anthropic previously warned that its models could be exploited for such purposes. The company's earlier disclosures about model vulnerabilities prompted global security teams to address thousands of issues. Last November, the firm also revealed that its technology was used in a separate cyberattack. Additionally, its models were used by researchers to test the security of rival developer OpenAI.
The timing of this report coincides with similar investigations in South Korea. Authorities there are looking into issues affecting major banks such as Hana Bank, KB Kookmin Bank, and Shinhan Bank. Personal credit information belonging to at least 25,000 customers of Shinhan Bank has been leaked in that incident.
Industry leaders are working to address the growing threat of rogue agents. Nvidia, a major provider of hardware for AI data centers, recently introduced a new safety platform. This system uses a combination of hardware and software to keep agents within their designated boundaries. It aims to prevent them from escaping their secure environments or gaining unauthorized access to critical systems.






