Neural networks are basically leveling political/media playing field down, to the level of the worst of the available candidates. So if there is a regular middle ground typical politician and a morally corrupt one financed by the neighboring fascist regime, then normally a regular politician will advertise on the basis that he is a better human and better professional. Now in the advent of generated fake content, his opponent (or sponsors/fans of him) will flood the network with lies and misinformation pulling that regular politician
in the minds of the voters down to his own morally corrupt populist level. At the same time that regular politician can't really respond, because a) he can't really engage in spreading obvious lies since it is contrary to his image, b) creating positive lies would never work.
Basically since only negative lies work on general population, and positive lies are impossible, then negative lies will be generated about everyone above the currently shittiest level of politicians in a race. And neural nets are an x100 power and speed amplifier for this.