My guess would be that it is because fake accounts are not merely fraud, but are often front organizations, either by lobbyists and political PACs or by corporation PR groups, or by nations states abroad and at home for their own covert 'PR'.
These fake accounts are quite difficult to detect.
The accounts that are easy to detect are avoided anyway, since users get to manage their own subscriptions. Some are even cherished for their novelty - horseebooks, etc.
Detecting these covert accounts means being a middle man for whichever party lays a claim about source - and this makes Twitter a middleman for other parties' curation.
Currently DIA uses internet signals collection, social media analysis and human intelligence to identify foreign messaging and popular messages that are harmful to national security.
They will submit requests to have these taken down (as the State Department did to Youtube to take down Al-Alwaki's videos) - though currently Twitter is not very cooperative.
I do not know how Twitter would handle a problem like this itself.
