Active Cryptojacking Campaign Targets GPUs via SEO Poisoning and AI Chatbots

Threat actors are running a coordinated cryptojacking campaign that has found a novel distribution channel: poisoning search engine results and manipulating AI chatbot responses to steer victims toward malware-laced downloads. The campaign, detailed by BleepingComputer, specifically targets machines with high-performance GPUs — workstations, render farms, ML training boxes, and any server with discrete graphics capacity.

What Happened

Attackers have built out a network of fake or compromised websites optimized to rank for technical queries — searches for software tools, drivers, utilities, and libraries commonly needed by developers and sysadmins. When a user lands on one of these poisoned pages, they're served a trojanized installer that deploys GPU mining malware alongside (or instead of) the expected software.

The more significant wrinkle: the campaign has also compromised AI chatbot recommendations. When users ask AI assistants for download links or software suggestions, some responses have been steering victims toward these malicious URLs. This is most likely the result of SEO-poisoned pages ranking well enough to surface when an AI assistant fetches live web results via retrieval-augmented search — meaning the attack works against any AI tool that performs real-time web lookups.

Why It Matters

The blast radius here is wider than a typical malware campaign. GPU-rich machines are ubiquitous in technical organizations: ML pipelines, GPU-accelerated CI runners, developer workstations, and rendering infrastructure. A cryptominer on any of these silently consumes compute, inflates cloud costs, degrades performance, and — critically — establishes a persistent foothold that can be upgraded to ransomware or data exfiltration at any time.

The AI chatbot vector is a genuine escalation that most security training hasn't addressed. If your team treats AI-generated download suggestions as implicitly trustworthy, that's an unpatched attack surface right now. SEO-poisoned sites look legitimate, filenames match what you searched for, and the installers often function normally — the malware simply comes along for the ride, making detection at install time unlikely without endpoint controls.

What to Do

Check GPU utilization immediately. Sustained, unexplained GPU load outside business hours is the clearest indicator. On Linux: nvidia-smi or rocm-smi. On Windows: Task Manager's GPU tab. Anything pegged high with no corresponding user process is suspicious.

Block known miner processes and pool destinations. XMRig, lolMiner, and T-Rex have well-known process signatures. Mining pool domains are broadly blocklisted in commercial threat feeds — make sure your DNS filtering and egress firewall rules include them.

Treat AI-sourced download links as untrusted. Any URL suggested by an AI assistant should be verified against the vendor's official domain before downloading. Add this explicitly to your team's security guidance.

Audit recent software installs on GPU-equipped machines. Review installs from the past 30–60 days, especially anything that arrived bundled with a legitimate tool or lacks a verifiable vendor signature.

Enforce application allowlisting or install controls. MDM, Group Policy, or endpoint controls that block unsigned executables will stop this class of attack even when a user intentionally downloads the payload.