The easy thing to do is to have two policies: one for politicians, and one for everyone else. The policy for politicians will be one that holds their actions to a much higher standard than everyone else. The policy can say that metrics will be used to determine if the statements by the politician are false or misleading, and if a certain threshold is reached, they can/will(/should!) be banned from the site. They can provide a detailed analysis about how they came to the conclusion to ban the user.