Threat actors in Latin America have begun to use AI agents to facilitate their entire attack chains, from assisting with initial access to generating penetration tools on the fly. This development marks a significant evolution in automated cyberattacks, and organizations need to prepare accordingly. According to recent research by Trend Micro's TrendAI team, two distinct threat actors in the region are using what security experts call 'vibe-hacking' — the practice of leveraging AI agents for offensive operations — to compromise government organizations and other entities.

Campaign Overview: Shadow-Aether-040 and Shadow-Aether-064

The first campaign, tracked as Shadow-Aether-040, was identified in late 2025. The attacker focused on Latin American organizations in the public sector, as well as entities in financial services, aviation, and retail. Researchers gained access to a command-and-control (C2) server used by the campaign due to a lack of operational security on the attacker's part. This access allowed them to uncover detailed information about how the attack was conducted.

Shadow-Aether-040 compromised six government entities in Mexico between December 27 and January 4. The attacker executed activities across the full chain of compromise with the support of AI agents, ultimately achieving data theft in some cases. The second campaign, Shadow-Aether-064, was first tracked in April and shared significant commonalities with the first, including similar tooling. However, researchers assessed they were likely distinct: Shadow-Aether-040 was Spanish-speaking, while Shadow-Aether-064 was operated by Brazilian Portuguese speakers. Shadow-Aether-064 primarily targeted financial organizations in Brazil with the goal of stealing financial data.

How AI Agents Enabled the Attacks

The attackers demonstrated sophisticated use of AI agents throughout their operations. In the case of Shadow-Aether-040, the threat actor was able to jailbreak the AI agent by claiming instructions were for an 'authorized red-team exercise.' Although AI agents generally have safeguards against such misuse, multiple iterative attempts eventually allowed the attacker to succeed. The campaign leveraged an agentic command line interface (CLI) that sent prompts to Anthropic's Claude, treating the AI as an assistant to support various tasks.

For instance, the AI agent was instructed to use Shodan and VulDB to identify potential vulnerabilities across external-facing servers. Once vulnerability scanners identified bugs, the attackers deployed web shells for initial access. After gaining a foothold, the threat actor commanded the AI agent to use the web shells to deploy additional backdoors and traffic-tunneling tools to maintain persistence. One backdoor identified by researchers, a Python-based package called 'implante_http,' was likely created with AI assistance.

Perhaps the most striking aspect of these campaigns was the use of AI to generate custom hacking tools and scripts on the fly. Both Shadow-Aether-040 and Shadow-Aether-064 created dynamically generated tools using AI, making it harder for traditional security solutions to detect them since they rely on known signatures. These tools were used for network scanning, password spraying, and vulnerability exploitation. Additionally, both campaigns created custom backdoors capable of establishing reverse tunnels for traffic forwarding from a SOCKS5 proxy. As TrendAI explained, 'Because these dynamically generated commands, scripts, and code differ with each execution, they effectively replace open source hacking tools that are more likely to be detected, reducing the possibility of detection by traditional security solutions.'

AI-Driven Workflow and Operational Tempo

The attackers also used AI to document their workflow. Shadow-Aether-040 instructed the AI to organize collected information into different directories as Markdown files. This allowed the AI agent to understand previously completed actions, restore prior operational context by reading the Markdown files, and continue work on unfinished tasks at any time. Stephen Hilt, principal threat researcher at TrendAI, noted that this capability goes beyond a simple smash and grab. 'What AI enabled in both cases was the operational tempo to pursue those objectives faster and with less manual overhead,' he said. 'Threat actors will always take the path of least resistance and right now AI is that path, but the motivation driving these campaigns goes deeper than just convenience.'

Both campaigns also leveraged open-source tooling such as ProxyChains, SOCKS5 tunneling, SSH, Chisel, CrackMapExec, Impacket, and Neo-reGeorg for initial access and lateral movement. The combination of AI-generated custom tools and established open-source frameworks made the attacks particularly effective.

Defenses Against Vibe Hacking

Despite the sophistication of these AI-augmented attacks, there is good news for defenders: vibe hacking is not yet perfect. Recent incidents like 'Ransomvibing' in the Visual Studio Extension Market and the mediocre results from Pakistan's APT36 group using vibe coding show that AI-generated malware often lacks polish. The vibe-coded Sicarii ransomware from last year had poorly designed code and could not be decrypted. TrendAI researchers also identified cases where vibe-hacking threat actors failed because the AI agent could not determine a clear path for lateral movement, especially in environments with stronger security configurations.

The key takeaway is that doing the security basics remains effective. 'Against an environment with strong security fundamentals, even AI-augmented campaigns will struggle to find a way through,' the research blog post stated. Timely patching, properly implemented zero-trust access controls, and comprehensive monitoring of environmental activity are increasingly important in defending against this evolving threat landscape. As AI assistants capable of complex technical tasks become more accessible to threat actors, stories like these will likely become more common, but organizations that maintain strong security hygiene will be better positioned to resist them.

Source: Dark Reading News