TL;DR EMPs probably won't work and simple magnets won't help on silicon machines. But offering honey-pot-counter-bots that attach to the offender and render them useless or use blood transfusion and a cleaning mechanism might work.
Professor Bradley Nelson of ETH Zürich managed to build a nanomachine that could be remote controlled and moved through a (hopefully dead) cows eye in 2004 Research site of the institute.
Raymond Kurzweil, director of engineering at Google, has published the book The singularity is near, which is quite some detailed reality check - with some restrictions (taken from one of the reviews)
This starts with the thesis: Technological change is exponential!
This has been true for many measures such as micro-processor size, cost of mass-produced goods, etc. It is not, however, a general rule of thumb to apply blindly to all things "technological"! This seems to be Kurzweil's big mistake. He extrapolates features of technology to an unrealistic infinity. [...]
By "remote controlled" and "stearing", they actually mean metal coated boxes (namely: "Nano Containers") that move along a persons veins. Professor Zhang Li from University of Hong Kong claims that he can remote control nanobots in a body. In fact it's nothing more than a magnet moving a box in a tube forwards or backwards. So they are far from real external controls as the electromagnets probably house the person in total.
the nickel-coated microbots are steered wirelessly by electromagnetic fields
Quite interesting are the mentions (incl. footnotes that I can't verify as I don't have the book at hand) of Professor Kensall Wise of the University of Michigan who seems to have built monitoring probes that allow precise monitoring of neural electrical activity and of Prof. Kazushi Ishiyama head of Ishiyama Laboratory of the Tohoku University who [acc. to R.Kurzweill] has built micromachines that deliver drugs directly to "precise locations in the brain". I would take this info with a grain of salt, as the university does not have any research paper about this.
According to Chris Phoenix, director of Research at the Center of responsible Nanoresearch, who has written about Nanofactories in his paper for the Journal of Evolution & Technology, the major problem seems to be mass manufactoring. In the paper he reviews the major previous ideas in the Background chapter showing the logic gaps in all the existing sketches and concepts.
Manufacturing While an optimistic view, like the one of Kurzweill states that if everything goes well, we will have usable nanomachines by 2020, there's no sign of mass production in the near future. Phoenix shows exactly how problematic design changes are when operating mass manufacturing at atom level. While building a prototype in a laboratory might already be possible, there's a lot of missing pieces to make this possible in a production chain.
Communication It may be possible that there are already prototypes (of which I couldn't find a paper) that could be remote controlled already, it won't be possible in near future to send a constant stream of visual real time data. The human eye is far beyond HD vision and is able to see up to 225 fps [Fighter pilots, no citation found] in a field of view, close to 180 degrees at around 550MP. Now try to transfer that amount of data and imagine how much energy you would need to transport that. So this is the most unrealistic part. Pretty much the same goes for audio data. Another problem will be that even when all the data can be read and transported LoFi, there would be constant interruptions through steel-concrete ceilings and walls, subway stations, etc. Also the range of field would be quite small and therefore would need a quite dense network of relays and signal amplifiers and relay stations.
Let's go through the list:
- ✓ Control your basic feelings (fear, joy, rage, ...)
- ✗ Check what you are seeing or hearing
- ✓ Cause pain on different levels
- ✓ Control your hormone levels
- ✓ Kill you
- ✗ They operate on the same substances your muscles are using
- ✓/✗ They can replicate inside your body (depending on the time this scenario is in)
- ✗ They have capabilities similar to nowadays cell phones (GPS, Wi-Fi, GPRS, 'background' offline programs, ...)
- ✓ They cannot read your mind.
Who is Who?
IEEE Spectrum has a nice PDF that shows every (in their view) important actor in this research field, allowing you to weight one opinion to the other more easily. Stay critical to everything you read as everything is just an assumption. Remember what we imagined the future to be in 2015? Flying rocket cars, holograms, etc. What have we got? Supercomputers in our pockets that we use to watch and share baby pictures and cat videos with people that we mostly don't meet in the real world. The future might no be what gets predicted by scientific researchers (in a single field).
Fighting off the Singularity
An interesting thought that goes against the assumptions made by transhumanists - who think that we will have better blood cells able to repair wounds quicker, transport more oxygen, fighting bacteria and viruses, etc. - is stated by Richard A.L. Jones in his article Rupturing the Nanotech Rapture
First, those building blocks--the cogs and gears made famous in countless simulations supporting the case for the singularity--have some questionable chemical properties. They are essentially molecular clusters with odd and special shapes, but it's far from clear that they represent stable arrangements of atoms that won't rearrange themselves spontaneously. These crystal lattices were designed using molecular modeling software, which works on the principle that if valences are satisfied and bonds aren't too distorted from their normal values, then the structures formed will be chemically stable. But this is a problematic assumption.
This means (more detail in the article) that the carefully crafted atomic structure of one molecular machine could be easily damaged if exposed to a substance it wasn't crafted for. So simply making a blood transfusion (as in exchange) with non-infected blood will give you a lot of chances to extract the machinery: Starting with mixing blood with water to destroy them and then using a simple filter mechanism to divide blood and water by its specific weight would be possible. It would as well be possible to craft your own nano machines that mimic target cells to attract an enemy machine and expose an atom on contact that attaches to the offending machine on contact rendering them useless or blocking their movement. Another possibility might be to use silicon magnets, as research by a team of scientists led by Paul Snijders of the Department of Energy's Oak Ridge National Laboratory and published in the New Journal of Physics in 2012, to extract nano machines built out of silicone atoms:
[...] The surprise is that while bulk silicon is non-magnetic, the edges of nano-ribbons of this material are magnetic. [...] the electron spins are ordered anti-ferromagnetically, which means they point up and down alternatingly. Configured this way, the up and down spin-polarized atoms serve as effective substitutes for conventional zeros and ones common to electron, or charge, current. [...] "By exploiting the electron spins arising from intrinsic broken bonds at gold-stabilized silicon surfaces, we were able to replace conventional electronically charged zeros and ones with spins pointing up and down".