Skip to main content

SATURDAY, 21 NOVEMBER 2026

VARINDIA
Breaking
Technology Disruption

AI-Powered Hackers Target Latin American Banks

3 min read18 views
Share:
AI-Powered Hackers Target Latin American Banks

The reported use of Claude, Grok and open-source AI models to automate cyberattacks against Latin American banks signals a significant shift in financial-sector security. The central concern is not simply that criminals are using AI, but that multiple models can potentially be combined into coordinated attack workflows requiring less continuous human intervention.

For banks, this raises questions about whether conventional security controls can respond quickly enough to machine-speed threats. The distinction between reported findings and independently established details is important, particularly when assessing the extent of automation and whether attacks resulted in financial losses.

Key Highlights

  1. 400+ financial targets: CyberCartel reportedly targeted banks and payment platforms across Latin America.

  2. Multi-model attacks: Claude, Grok and Moonshot AI models supported different operational functions.

  3. Autonomous coordination: AI agents reportedly handled intelligence gathering, strategic analysis and attack orchestration.

  4. Sensitive data exposure: Researchers identified indications of compromised credentials, customer information and account access.

  5. AI safeguards challenged: Anthropic blocked identified misuse after receiving Straiker's findings.

  6. Machine-speed cybercrime: Automation could increase attack volume while reducing operational costs for criminals.

  7. BFSI security imperative: Banks need continuous authentication, behavioural monitoring, stronger API protection and AI-powered incident response.

Cybercriminals are moving beyond conventional hacking tools, using artificial intelligence to coordinate sophisticated attacks against financial institutions with minimal human intervention.

According to an investigation by cybersecurity startup Straiker, a hacking group identified as CyberCartel deployed interconnected AI agents against more than 400 financial targets across Latin America, including banks and payment platforms.

The campaign reportedly combined Anthropic's Claude, xAI's Grok and models developed by Chinese AI company Moonshot AI, demonstrating how attackers can exploit multiple AI ecosystems within coordinated operations.

Grok-powered agents reportedly gathered intelligence about potential targets, while Claude Opus and Sonnet agents analysed attack strategies. Another Claude instance operating through Cursor coordinated activities across the network.

Straiker's investigation suggests the attackers obtained passwords, accessed customer information and entered banking accounts with limited human involvement. However, the full financial impact remains publicly unconfirmed.

Researchers uncovered the operation after CyberCartel inadvertently exposed information about its internal activities, providing unusual visibility into the group's automation capabilities.

Anthropic reportedly investigated the findings and blocked the group's access to Claude. Nevertheless, restricting individual models may not eliminate threats involving interchangeable commercial and open-source systems.

The strategic concern is the industrialisation of cybercrime. Autonomous agents can accelerate reconnaissance, coordinate intrusion attempts and repeatedly exploit weaknesses faster than conventional security teams can investigate incidents.

For Indian banks, the findings reinforce the importance of continuous identity verification, behavioural analytics, API security, privileged-access controls and AI-driven threat detection across digital banking infrastructure.

The emerging cybersecurity challenge is no longer simply human hackers using AI assistance. It is the possibility of interconnected autonomous agents executing sustained attacks at machine speed. Financial institutions must respond with equally adaptive, continuously monitored security architectures capable of detecting malicious behaviour before financial losses escalate.

Join the Conversation

Loading comments…

Share your thoughts on this story. Comments are moderated before they appear.

Your email address will not be published

0/2000

Be respectful. Comments are reviewed before publication, and spam or abusive language will be removed.