Independent Apep Bar reporting and analysis. Source-derived facts are attributed; where a supplied video could not be independently reviewed, its title is treated only as a starting point.
Attackers may no longer need to do every step themselves
An October 2026 investigation reported that a group called CyberCartel used AI agents to automate aspects of cyberattacks against hundreds of Latin American financial targets, citing research from the cybersecurity firm Straiker. Researchers alleged the tools assisted with targeting, stolen credentials, and access to some accounts.
An agent is a tool, not an independent villain
AI agents can chain routine actions and react to results with reduced human intervention. Used offensively, that can make ordinary criminal techniques faster or more scalable. The term ‘AI hacking a bank’ risks implying a sentient machine independently chose a target; the reported operations involved human attackers using software.
What defenders should ask
The security questions are how much access was actually gained, which steps were automated, whether evidence is reproducible, and what defenses interrupted the attacks. The available reporting should not be confused with a claim that all banks have been breached.
Apep Bar asks
When software automates cybercrime, is the bigger danger a new attack method—or the falling cost of conducting old attacks?
References and further reading
Featured image is illustrative and does not purport to show the event, individual, or equipment discussed unless explicitly stated.
