Introduction to Canto Incognito

Recent findings from cybersecurity analysts have unveiled a new strain of malware, dubbed Canto Incognito, that targets publicly accessible artificial intelligence (AI) systems and large language models (LLMs). The primary objective of this malicious software is to deploy cryptocurrency miners, thereby enhancing the scale of a burgeoning botnet.

Mechanics of the Malware

The Canto Incognito campaign integrates sophisticated methods to achieve its goals. It is known to install various cryptocurrency mining software, such as XMRig and Iron, and link affected devices to Kryptex, a Russian-based cryptocurrency mining service.

As noted in a report by Lumen Black Lotus Labs, “Compromised hosts are reused to expand the botnet,” indicating that infected servers may be repurposed to locate and exploit other vulnerable systems.

Innovative Command-and-Control Techniques

The malware is assigned the codename PoeLLM, which reflects its clever technique of hiding command-and-control (C2) addresses within poetic verses uploaded to a GitHub repository. As described by Ryan English, an information security engineer at Lumen Technologies, “Each time they set up a new C2, they change a few words in the poem, and the malware derives the address from the key associated with those words.” The repository first appeared on April 13, 2026.

Target Vulnerabilities

Canto Incognito primarily targets enterprise-level services with internet exposure, including deployments such as LiteLLM, Gotenberg, and platforms like Gitea and Ivanti Sentry. The aim is to exploit their computational capabilities for illicit mining activities.

As of now, over 3,400 victim servers have been identified, with a significant concentration of infections in the United States and Western Europe. The malware has reportedly been active since April 2026.

Operational Scale and Trends

According to Lumen, the malware campaign reached its peak in mid-June, with approximately 2,200 servers compromised at that time and around 800 active on a daily basis. There are indications of evolving tactics, with increased activity around SSH and login portals hinting at ongoing experimentation with distributed brute-force attacks.

Exploitation and Expansion Tactics

Notably, infected systems are also utilized to scan the internet for similar vulnerable targets, facilitating the spread of the malware further afield. The campaign seeks to capitalize on known vulnerabilities in publicly accessible services, effectively transforming them into a mining botnet while continually scouting for additional targets.

Lumen Black Lotus Labs has linked the campaign to an Italian-speaking threat actor based on linguistic evidence and netflow analysis, underscoring the global reach of the threat.

Conclusion: The Growing Threat of AI Targeting

The increase in attacks on AI infrastructure is alarming. As Lumen emphasizes, “AI infrastructure is becoming an attractive target.” These exposed AI and LLM services are valuable not just for their software vulnerabilities, but also due to the valuable data they harbor and their powerful hardware capable of extensive mining operations.