A new [article](http://www.politico.com/magazine/story/2017/09/08/how-facebook-changed-the-spy-game-215587) was released in which a former FBI counter intelligence agent detailed the current situation with regard to foreign propaganda being spread not by human agents but by bots. In other words, AI now can be used to produce fake stories, fake social media activity, do astroturfing, or to create and generate all sorts of confusion. This is something the FBI has a mandate to counter. The quote below details it:

>The vast majority of counterintelligence cases I worked in the FBI involved a foreign intelligence service (FIS) conducting what we called “perception management campaigns.” Perception management, broadly defined, includes any activity that is designed to shape American opinion and policy in ways favorable to the FIS home country. Some perception management operations can involve aggressive tactics like infiltrating and spying on dissident groups (and even intimidating them), or trying to directly influence U.S. policy by targeting politicians under the guise of a legitimate lobbying group. But perception management operations also include more passive tactics like using media to spread government propaganda—and these are the most difficult for the FBI to investigate.

As I've said in other posts, public opinion is extremely important. Public opinion is how people know what behaviors to adopt to be perceived as "good" or as "bad" or to be perceived as ethical. Social media can give us a distorted idea of where public opinion really is and this could cause us to adopt behaviors which we think conform to public opinion but which do not. The FBI is looking at the fact that foreign nations have every incentive to want US citizens (potential voters) to think favorably of their nation. There are also some foreign nations who may have the incentive to want to make US citizens hate their own country, or think negatively about their own government.  The power of artificial intelligence is being under-estimated here and I think these examples highlight the danger of leaving AI centralized into silos because then only a handful of companies might have the power to shape public opinion.

> Still, the FBI has a legal duty to stop, or “neutralize,” foreign intelligence operations occurring on U.S. soil. In the case of propaganda operations, one effective way to do this while avoiding First Amendment land mines is to approach the FIS agent directly. There isn’t exactly a law that prevents an agent from being a “source” for a journalist. But there is a law called the Foreign Agent Registration Act (FARA) that requires any individual acting on behalf of foreign interests to register as a foreign agent with the State Department.Still, the FBI has a legal duty to stop, or “neutralize,” foreign intelligence operations occurring on U.S. soil. In the case of propaganda operations, one effective way to do this while avoiding First Amendment land mines is to approach the FIS agent directly. There isn’t exactly a law that prevents an agent from being a “source” for a journalist. But there is a law called the Foreign Agent Registration Act (FARA) that requires any individual acting on behalf of foreign interests to register as a foreign agent with the State Department.

>  But Congress could pass legislation that requires social media companies to cooperate with counterintelligence in the same ways they do with law enforcement. For example, the Communications Assistance for Law Enforcement Act (CALEA) requires telecommunications companies to design their digital networks in such a way that would permit wiretaps for criminal cases. 

Of course we would have to question what would this cooperation mean? Counterintelligence at times focuses on foreign agents and foreign operations but what about situations where counterintelligence focuses it's attention on American citizens, some of whom might actually be innocent or merely guilty by association? I would think this is an area where there are dangers and creating a new law is risky.

> Another option is to borrow from the counterterrorism toolbox and designate entities known to be acting on behalf of foreign intelligence—like Russia’s “troll farm,” the Internet Research Agency—as nonstate intelligence services, the same way we now designate some nonstate actors as foreign terrorist organizations. 

What exactly is a non-state intelligence service and how could this become a problem for innocent people? Again while the FBI has a mandate to conduct their mission, at the same time how would it be distinguished which people are legitimately and willingly part of the troll army vs people being trolled and manipulated by disinformation to believe in the propaganda? I don't see how it would be so clear cut.  AI also is likely going to automate a lot of the propaganda generation so the troll army might just be software.

References
---
1. http://www.politico.com/magazine/story/2017/09/08/how-facebook-changed-the-spy-game-215587