This Go Botnet Is Hunting AI Services and Stashing Cloud Keys

Emily Davis · 2026-07-17

A new botnet called NadMesh has been spotted in the wild since early July, and it's got a very specific target: exposed AI services. The operator's own dashboard claims it has already collected 3,811 unique AWS keys. That's a lot of cloud credentials in a short time. ### What Is NadMesh Targeting? NadMesh is written in Go, which makes it fast and hard to detect. It uses a Shodan harvester to keep its scan queue full of popular AI tools that teams often stand up quickly but forget to secure. The list includes: - ComfyUI - Ollama - n8n - Open WebUI - Langflow - Gradio These are the image generators, local model runners, and workflow builders that developers love for their speed and flexibility. The problem? They're often deployed with default settings and open ports, making them easy pickings for a botnet like NadMesh. ### How the Botnet Works The botnet scans the internet for these services, looking for ones that are exposed without proper authentication. Once it finds a vulnerable instance, it tries to grab cloud keys and Kubernetes tokens from the environment. The harvested data is then sent back to the operator's dashboard, which reportedly shows 3,811 unique AWS keys so far. > "The intel feed behind that counter is a goldmine for attackers," says Emily Davis, Head of Digital Privacy at Antidetectbrowsershub. "If you're running any of these tools without a firewall or authentication, you're basically handing over your cloud access." ### Why This Matters for AI Teams AI development moves fast. Teams spin up ComfyUI for image generation or Ollama for local models in minutes, often on cloud instances with public IPs. Security takes a back seat to speed. But NadMesh shows exactly why that's dangerous. A single exposed service can leak AWS keys, Kubernetes tokens, and other credentials that give attackers full access to your cloud infrastructure. ### How to Protect Yourself Here are some practical steps to keep your AI services safe from botnets like NadMesh: - **Use a firewall** to restrict access to your AI tools. Only allow traffic from trusted IPs. - **Enable authentication** on all services. Default settings often have no password. - **Scan your own infrastructure** with Shodan or similar tools to see what's exposed. - **Monitor for unusual activity** like unexpected API calls or new keys being created. - **Rotate cloud keys regularly** and use short-lived tokens where possible. ### The Bigger Picture NadMesh isn't the first botnet to target AI services, and it won't be the last. As more teams adopt tools like ComfyUI and Ollama, attackers will keep looking for easy targets. The key takeaway? Don't assume your services are safe just because they're internal or temporary. If it's online, it can be found. ### Final Thoughts If you're running any of these tools, take a few minutes to check your security settings. It's a small investment that could save you from a major breach. And if you're using an antidetect browser to manage multiple accounts or protect your identity, remember that the same principles apply: keep your tools updated, use strong passwords, and don't expose anything you don't need to. Stay safe out there.