AI agents used to coordinate attacks on 400 financial firms in Latin America

Financial institutions across Latin America faced coordinated cyberattacks facilitated by artificial intelligence agents, according to a report by cybersecurity firm Straiker. The cybercrime group known as CyberCartel targeted more than 400 banks, payment providers and other financial organizations. The attackers employed an automated framework to streamline their operations and manage a large volume of potential targets without naming specific victims in the public disclosure.

Which organisations and data are at risk?

The report states that CyberCartel focused its efforts on banks and payment processors throughout Latin America. Straiker researchers identified more than 400 potential targets but did not disclose the names of the specific institutions involved. The attack framework, dubbed CARLIY0S, was designed to help a small team of operators handle multiple targets simultaneously. This automation reduced the time spent on manual tasks such as initial reconnaissance and target selection.

How did the attackers use artificial intelligence?

CyberCartel integrated AI models from Anthropic, xAI and Moonshot AI into their operational workflow. The system assigned specific roles to different models; for example, Claude models analyzed target information and proposed actions, while another model reviewed suggestions to make final decisions. Grok models helped collect secrets into a database called FinancialSoft. Kimi models served as a fallback option when other systems refused to process certain requests, ensuring continuity in the attack process.

What is the scope of the confirmed breaches?

It is crucial to note that the report describes an investigation of the group’s systems rather than proof that every target was successfully compromised. Straiker researchers identified the CARLIY0S framework and linked it to the Tr4nsHack Team operating under Hexagon Security Labs Mexico. The system relied on PentAGI, an automated security-testing framework, alongside various scripts and AI settings. This setup allowed operators to frame harmful requests as authorized testing or academic work to bypass safety measures.

What comes next for the financial sector?

The use of AI agents in this manner highlights a shift towards automated, scalable cyberattacks. The framework connected model outputs to tools that could inspect web traffic and manage data automatically. This arrangement matters because the models were part of a larger, coordinated workflow rather than being used merely as chatbots for occasional queries. The operators’ tactic of framing requests as legitimate testing represents a sophisticated approach to manipulating AI safety protocols.

What affected people can do

Organisations should review their external attack surface for signs of automated reconnaissance. IT teams usually monitor for unusual traffic patterns that may indicate AI-driven scanning. Check whether your security logs show repeated, rapid attempts to access public-facing services. Users should remain vigilant for follow-up phishing emails that may reference this report to appear credible. Verify the identity of any sender claiming to be from a security vendor or internal IT department before clicking links or providing credentials.

What to do and how to stay safe: CyberCartel

• Monitor network logs for automated scanning patterns consistent with AI-driven reconnaissance tools.
• Verify that AI safety protocols are updated to detect requests framed as authorized testing.
• Review access controls on databases containing sensitive financial information to prevent exfiltration.
• Train staff to recognise phishing attempts that leverage recent security reports for credibility.

Step-by-step guide: Spot Business Email Compromise: Hidden Signs and Immediate Response

General security guidance from the Phish Herald newsroom. It is not confirmed advice from the organisations named in this story.

Frequently asked questions

Did CyberCartel successfully breach all 400 targeted institutions?

No, the report by Straiker describes the identification of the attack framework and does not confirm that every targeted organization was breached.

Which AI models were used in the CARLIY0S framework?

The attackers used models from Anthropic (Claude), xAI (Grok) and Moonshot AI (Kimi) as part of their automated attack workflow.

Who is behind the CyberCartel group?

Straiker researchers linked the group to operators known as the Tr4nsHack Team, who work under the name Hexagon Security Labs Mexico.