If you use languages in your thoughts or when reasoning with your self, you will produce a lot of subliminal speech that can be picked up from the nerves. By placing neuroprosthetics close to the cranial nerves responsible for producing speech, reading subliminal speech is possible.
The vocalization controller (sound ROM and programmed arithmetic logic) and speech output servo interface is what is being discussed here, not the main processor, processing queues (RAM) programmed and/or physical memory.
Creating a composition is an entirely separate host of groups of processes and sub processes from that of "printing" one that has been formed and queued for output.
Do your lips move when you listen, read or think?
Maybe with enough electrodes synced appropriately and the proprietary formatting algorithms (not to mention filetype codecs) one could some day read-out somebody's whole "hard drive" but that's far from what this silly hardwired "spinal tap" thingy is, or appears to be sort of capable of.
The most sophisticated prosthetic interfaces to date are merely for optical input. Nothing is anywhere near cracking the OS