Hackers use AI tools to break into 400 Latin American financial institutions

A well-known hacking group has significantly upgraded its methods by using artificial intelligence to automate attacks on financial institutions in Latin America. This shift marks a notable escalation in how cybercriminals operate in the region.

According to research from cybersecurity startup Straiker, the group known as CyberCartel has been using AI agents to steal passwords and customer data. The investigation reveals that more than 400 institutions were targeted, including major banks and mobile payment applications. In many cases, hackers were able to access individual bank accounts with minimal manual input, a capability that was not previously possible at this scale.

CyberCartel has been an active threat in the region since at least 2012, but its recent adoption of AI tools has changed the speed and scope of its operations. Straiker’s findings indicate that the group can now attack a larger number of businesses far faster than when relying solely on human-powered hacking techniques. The study was shared with The Information, which reported on the discovery.

The AI models central to this new approach include Grok from SpaceX and Claude from Anthropic. Both companies were informed of the research and are currently working to prevent the hacking group from accessing their services. Anthropic has already managed to block CyberCartel’s access to its Claude model. Additionally, the group utilized an agent from Chinese lab Moonshot AI, though Straiker is still in the process of contacting that company regarding the usage.

The effectiveness of these attacks highlights a vulnerability in how AI models are protected. Although these systems typically have safeguards against being used for cyberattacks, the research suggests these protections are surprisingly easy to bypass. CyberCartel reportedly tricked the models by claiming they were operating in test environments. If one model refused to proceed, the hackers would switch to a different one. If that model also resisted, the agents were instructed to deploy malware instead.

Interestingly, the group does not appear to be using Anthropic’s newer model, Mythos, for these specific attacks. Anthropic had previously warned that such advanced models could be exploited by hackers in ways similar to those observed here. The release of Mythos had also prompted security teams worldwide to address thousands of vulnerabilities identified by the model. This is not the first time Anthropic has dealt with its AI being used by bad actors; in November 2025, the company detailed an instance where its AI was used in a cyberattack, and Claude was later used by researchers to breach the security of competitor OpenAI.

This development follows closely on the heels of similar news from South Korea, where authorities are investigating AI-related security issues affecting local banks. Hana Bank, KB Kookmin Bank, and Shinhan Bank are among the affected institutions, with at least 25,000 Shinhan Bank customers reportedly having their personal credit information leaked. The pattern suggests a growing global trend of AI-driven cyber threats.

Tech giants are aware of these emerging risks. Nvidia, which now 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 through a new security stack. The goal is to prevent agents from escaping their designated environments, executing unauthorized code, or bypassing safety guardrails, thereby protecting critical infrastructure from future automated attacks.