What assurances do we have that this system will not be used to flag "extremist content" i.e. memes and to send that information to law-enforcement in the future?