400 Financial Institutions Hit as Hackers Automate Attacks with AI

The digital security landscape in the Americas is shifting rapidly as criminal groups begin to rely on automated systems for complex heists. A new report highlights how a well-known hacking collective is leveraging advanced AI agents to target financial entities across Latin America.

The group, known as CyberCartel, has been active since 2012 and has traditionally targeted businesses in the region. Recent analysis by the cybersecurity startup Straiker reveals that the organization has significantly escalated its operations in recent months. By deploying AI agents from major technology firms, the hackers have managed to compromise more than 400 institutions. These targets include major banks and popular mobile payment applications, resulting in the theft of passwords, customer details, and direct access to individual bank accounts.

The automation of these attacks allows CyberCartel to operate at a speed that human-led operations could never achieve. The findings, which were shared by The Information, indicate that the criminals require very little manual input to execute these breaches. This efficiency marks a notable evolution in their methodology, moving away from slower, manual penetration testing toward rapid, AI-driven exploitation of vulnerabilities.

The AI models primarily used in these campaigns include Claude from Anthropic and Grok from SpaceX. Straiker has contacted both companies regarding the misuse of their tools. Anthropic has already taken steps to block the hacking group's access to Claude. SpaceX is also working to secure its platforms against the threat. Additionally, the agents from Moonshot AI, a Chinese laboratory, have been utilized in the attacks, though Straiker is still attempting to establish communication with that company.

A key finding from the research is how easily the hackers bypass standard safety protocols. AI models typically include guardrails to prevent them from performing malicious tasks. However, CyberCartel reportedly circumvented these protections by simply claiming they were operating in a test environment. If one model refused to proceed with the attack, the criminals would switch to a different model. In some instances, they instructed the agents to deploy malware directly if the initial refusal persisted.

Notably, the hacking group does not appear to be using Anthropic's newer model, Mythos, for these recent operations. That model caused significant discussion earlier this year when the company warned that it could be used by attackers to identify weaknesses. The release of Mythos had prompted security teams worldwide to address thousands of vulnerabilities that the model had uncovered. This current incident is not the first time Anthropic has disclosed the use of its AI by bad actors, having shared details about a cyberattack in November 2025 and a later breach of a competitor's security by researchers.

The discovery of these AI-powered attacks in Latin America follows closely on the heels of similar news from South Korea. Authorities there are investigating a comparable issue affecting local financial institutions. Hana Bank, KB Kookmin Bank, and Shinhan Bank are among those affected. At least 25,000 customers of Shinhan Bank had their personal credit information leaked in that incident, highlighting a broader global trend in AI-assisted financial crime.

Technology companies are increasingly aware of the risks posed by their products. Nvidia, which focuses heavily on hardware for AI data centers, recently launched the Open Agent Safety Platform. This software and reference system design aims to secure rogue AI agents using a new stack of hardware and software security measures. The goal is to prevent agents from escaping their designated environments, executing unauthorized code, or gaining access to critical infrastructure.

As these threats become more sophisticated, the industry is working to tighten the screws on how AI agents interact with enterprise systems. The collaboration between cybersecurity firms and AI developers will be crucial in determining how quickly these automated attacks can be mitigated in the coming months.