A simple example will suffice.
Let's say I own a digital camera, or a device that contains a digital camera, that uses encrypted firmware.
Now let's say that this firmware has been written in such way that, at regular intervals, and without any form of notification to the user (like a LED blinking or a tell-tale shutter noise), it takes low resolution pictures of what it sees, and stores these unrequited pictures into its embedded storage (which too would be encrypted).
Then, when the user takes a normal high res picture using the digital camera, the firmware adds "noise" to it that contains encrypted image(s) from the hidden low-res pictures it has in memory.
With hi-res image files in the MB range or even better: digital video, it has become exceedingly easy to add "meaningful noise" and play a little steganography on digital photography. There are also ways to ensure sure that an useful data payload can still be recovered, even if the user downsizes the images.
Blissfully unaware of this, the user then uploads some of these high res picture to a public website, as one does. Then malicious entities/government agencies can just run a search on EXIF (if the user didn't remove this data, but even then, with the proper resources, parsing images all day to check for known steganographic payload is not that big a feat) and spy on you at length, provided you keep sharing pictures with friends, etc, and without your knowledge...
Of course, this kind of far fetched scenario would never happen... just like printer manufacturers would never add hidden marks to every page printed, that would uniquely identify what printer (and by extension who) printed some data.