Skip to content
PoeLLM Malware Compromises 3,400+ AI Servers via GitHub Poem

PoeLLM Malware Compromises 3,400+ AI Servers via GitHub Poem

First seen 7 Oct 2026, 17:28 UTC • •

Article Content

Browse articles
ThreatCluster AI
ThreatCluster •October 8, 2026 at 16:42 UTC

The PoeLLM malware campaign, tracked by Lumen Technologies' Black Lotus Labs, has infected over 3,400 servers since April 2026, primarily targeting exposed AI services like LiteLLM and Ollama. The malware cleverly hides its command-and-control (C2) server addresses within a poem posted on GitHub, allowing attackers to rotate infrastructure with minimal detection. The campaign, dubbed Canto Incognito, is believed to be financially motivated, utilizing compromised servers for cryptocurrency mining and scanning for additional victims. Key vulnerabilities exploited include CVE-2026-42271 in LiteLLM and CVE-2026-10520 in Ivanti Sentry. The operation has seen peak activity of nearly 800 infected servers per day, with a concentration of victims in the U.S. and Western Europe. Researchers have noted that the malware can also enable remote code execution, raising concerns about further exploitation.

Start a free Starter trial for enhanced analysis

Ask AI about this cluster

Updated 44d ago How this analysis works

Timeline

2026-04-13
First GitHub commit of the malware poem
The initial commit containing the poem that conceals the C2 address was made on GitHub.
Shattered
2026-05-08
CVE-2026-42271 published
A command injection vulnerability in LiteLLM was published, enabling exploitation by malware.
BleepingComputer
2026-06-08
CVE-2026-42271 added to CISA KEV
CISA added the CVE to its Known Exploited Vulnerabilities catalog due to active exploitation.
Helpnetsecurity
2026-06-09
CVE-2026-10520 published
A critical vulnerability in Ivanti Sentry was published, further enabling malware exploitation.
BleepingComputer
2026-06-11
CVE-2026-10520 added to CISA KEV
CISA added the critical CVE to its KEV catalog due to confirmed exploitation in the wild.
Helpnetsecurity
2026-10-07
PoeLLM malware report published
Black Lotus Labs reported the scale and methods of the PoeLLM malware campaign, confirming over 3,400 infected servers.
Cyberscoop

More articles in this cluster (24)

Following this threat?

Track PoeLLM, Black Lotus Labs and CVE-2026-10520 in your own feed — you're alerted when they show up in new reporting, leak sites or exploitation.

Free account · no card needed

Common questions

Which systems are affected by PoeLLM?
The malware primarily targets exposed AI services like LiteLLM, Ollama, Gotenberg, and Gitea.
Is there active exploitation of the vulnerabilities?
Yes, both CVE-2026-42271 and CVE-2026-10520 are actively exploited in the wild.
What should organizations do to protect against PoeLLM?
Organizations should apply security updates for affected systems and reduce public exposure of critical assets.