CyberCartel Automates Bank Heists with AI Agents

A notorious hacking group is using artificial intelligence to steal bank credentials at a pace that traditional human teams simply cannot match. CyberCartel, a crew with a history of targeting businesses in Latin America, has integrated AI agents into its operations to automate attacks on financial institutions. This shift marks a significant escalation in how the group operates, allowing them to breach more systems with less manual effort from the criminals themselves.

The group has been active since at least 2012, but its recent adoption of AI has changed the landscape of its threats. For years, hackers relied on manual processes to find and exploit vulnerabilities. Now, they deploy software agents to handle repetitive tasks, such as testing passwords and accessing customer data. This automation enables the group to attack a higher volume of targets simultaneously, a capability that was previously limited by the speed of human researchers.

CyberCartel is primarily using models from Anthropic and SpaceX to power these operations. Anthropic provides the Claude model, while SpaceX supplies the Grok agent. A startup called Straiker, which shared its findings with The Information, identified these tools in the group's attack chain. Straiker notified both companies about the usage, and both are now working to restrict the group's access to their services. Anthropic has already blocked CyberCartel from using Claude, though the team is still trying to reach out to Moonshot AI, a Chinese lab whose agents are also being used.

The hackers have found ways to bypass the safety protections built into these AI models. Typically, these systems are designed to prevent AI from carrying out cyberattacks. However, 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 tasked with deploying malware instead, effectively forcing the AI to act as a tool for intrusion.

Interestingly, the group does not appear to be using Anthropic's Mythos model for these recent attacks. Mythos made headlines earlier this year when the company warned that hackers could use it for similar purposes. The release of Mythos had prompted security teams worldwide to address thousands of vulnerabilities that the model had identified. Despite this potential, CyberCartel has stuck with Claude and Grok for its current campaign.

This is not the first time Anthropic has discovered its technology being used by bad actors. In November 2025, the company shared details on how its AI was used in a cyberattack. Subsequently, Claude was used to break the security of OpenAI, a major competitor in the AI space. While that specific incident was carried out by researchers rather than criminals, it highlighted the growing risk of AI models being turned against the very companies that create them.

The discovery of CyberCartel's AI-powered attacks comes just days after news broke that South Korean authorities are investigating a similar issue affecting local banks. Hana Bank, KB Kookmin Bank, and Shinhan Bank are among the institutions affected in South Korea. At least 25,000 Shinhan Bank customers have had their personal credit information leaked, suggesting that the trend of AI-assisted banking theft is not limited to Latin America.

AI companies are increasingly aware of the risks their products pose to the digital ecosystem. Nvidia, which now focuses on producing hardware for AI data centers, recently launched the Open Agent Safety Platform. This software platform and reference system design are intended to secure rogue AI agents through a new hardware and software security stack. The goal is to prevent agents from escaping their sandboxes, executing unauthorized code, or gaining access to critical infrastructure, a problem that has become more urgent as AI plays a larger role in cybercrime.

As AI agents become more sophisticated, the line between human-led and machine-led attacks continues to blur. The case of CyberCartel demonstrates that criminals are rapidly adapting to new technologies. For financial institutions, the challenge is no longer just securing against known human behaviors, but also protecting against autonomous software that can adapt, bypass guards, and scale its attacks in real-time. The industry is watching closely to see if new safety standards can keep pace with the evolving tactics of these digital criminals.