AI-enabled NIK C2 framework streams keystrokes and runs OpenAI prompts on victims
The claim
The authors assert that NIK C2 is a fully functional AI-integrated command-and-control framework that collapses collection and exploitation into a single continuous session.
Published as: “NIK C2: The new AI-enabled command-and-control framework”
So what
Defenders must monitor outbound OpenAI API usage and continuous behavioral indicators rather than relying on traditional offline analysis windows.
The five-minute version
Cisco Talos observed multiple Go-based Windows implants called NIK C2 that embed an OpenAI API client, allowing attackers to send arbitrary prompts to ChatGPT from the victim host. The implant also streams keystrokes, captures screenshots, and can deploy a screen-locking ransomware module without encrypting files. The authors provide detailed reverse-engineered evidence of dual-channel TLS-WebSocket C2, AES-256-GCM encryption, and a 52-command surface, and claim the framework is fully operational in the wild.
What they found 2
- Ransomware module locks the screen without encrypting files, leaving no forensic trail
“The design of this ransomware module is notable for several reasons. It does not encrypt any files, which means it leaves no forensic trail of file modification timestamps or encryption markers.”
- Implant routes traffic through three prioritized C2 tiers to survive takedowns
“To survive takedowns, the implant routes traffic across three prioritized C2 tiers, which are runtime-updated URLs, an embedded encrypted server list, and a hardcoded fallback address.”
What undercuts it the paper’s own results
- We assessed with moderate confidence that NIK C2 is not merely a proof of concept, but a fully functional, AI-integrated C2 framework.
Stated limits 3
- All binaries observed were compiled with Go 1.23.4, placing the sample set no earlier than late 2024.
- The implant uses a hardcoded seed for AES-256-GCM encryption, which may be a weak point for defenders.
- The authors state their confidence as "moderate" rather than definitive.
What’s new here 4
- First public report of malware embedding live OpenAI API calls on compromised hosts
- Detailed description of a screen-locking ransomware that avoids file encryption
- Analysis of a dual-channel C2 architecture with hard-coded AES keys and three-tier URL fallback
- Evidence of credential dumping, Wi-Fi password theft, and bootkit preparation in a single implant
Artifacts & methods 5
- NIK C2 implantsystem
- OpenAI Chat Completions API clientprimitive
- screen-locking ransomware modulesystem
- keyloggerprimitive
- dual-channel TLS WebSocket C2system