Home / Blog / Exposed AI Agents Are the New Shadow…
AI and LLM Security
Exposed AI Agents Are the New Shadow IT – Except They Can Act
175,000 Ollama servers exposed to the internet, 8,000 MCP servers wide open, and 31% of those examined answering with no authentication at all. Same film, new cast - with one twist.
Everybody is rushing to deploy AI agents. Almost nobody is securing them.
This week a tool called AIMap was published that scans the internet looking for exposed AI servers. The results? Over 175,000 Ollama servers exposed to the world. Over 8,000 MCP servers completely open. A million AI services scanned, a large proportion of them with no authentication whatsoever.
Let's stop there for a second. Of 5,200 Ollama servers examined in depth, 31% responded to requests without requiring any identification at all. 518 of them wrapped models from Anthropic, OpenAI and Google - meaning somebody is paying for an expensive API, deploying it on a server, and not even putting a password on it.
This is not a bug. It is a pattern.
Every new technology goes through the same cycle
Somebody invents something remarkable. Everybody rushes to deploy. Security? We'll handle that later. And later never arrives.
In 15 years of pentesting I have seen this with everything. With APIs that were open to the world. With Jenkins servers with no password. With S3 buckets anybody could read. Now it is AI agents. Same film, new cast.
But this time there is a twist
These AI agents are not just an endpoint returning JSON. They have access to tools. They have permissions. They have system prompts that leak out and reveal the whole internal logic. Some run as root. Some carry credentials embedded in the code. Some can execute code with no sandbox at all.
This is not shadow IT. This is shadow IT with the ability to act on the organisation's behalf.
I am not saying do not deploy AI agents. I am saying that anyone deploying an agent with no auth, no TLS and no limit on its permissions is simply opening a back door into their own organisation and inviting everyone in.
So tell me: who in your organisation is responsible for the security of the AI agents? If there is no clear answer, you have a problem.
I first shared a version of this as a LinkedIn post on 2026-05-17. It is republished here, lightly edited, so it is easier to find and reference. — Erez Metula
Keep reading
More from the blog
AI and LLM Security
Somebody Wired the Darknet Into Your AI. What Could Go Wrong?
One MCP server packing 66 tools, feeding onion-site content straight into your model's context. The darknet is the most hostile input that…
Read itAI and LLM Security
Every Week Someone Asks Me When AI Will Replace Pentesters
Datadog released a scanner that sends your code to an LLM instead of matching patterns. After 20 years in this field: the…
Read itAI and LLM Security
Your AI Coding Tool Walks Straight Past Your Defences
A prompt injection hidden in a README, and the tool writes to your zshenv. Every new shell runs the attacker's code. And…
Read itTell us what the system does and what worries you.
If a penetration test is not what you need yet, we will say so.