Robinson Meyer at the Atlantic reminds us that Twitter carves out big exceptions to its new policy.

The guidelines do not draw a distinction between user behavior on or off the site: If someone tweets only in coded language on Twitter, but calls for racial violence or genocide elsewhere on the web or in person, then they could still be banned from the service.

While logos or symbols affiliated with hate groups will not result in someone getting banned, they will carry a sensitive media tag, meaning that they will not automatically display to the site’s users.

But “context matters when evaluating for abusive behavior,” warns Twitter, and they have included two big exceptions in the new policy. First, their ban on advocating violence against civilians does not apply to “military or government entities.” Second, they may moderate their own rules if “the behavior is newsworthy and in the legitimate public interest.”

Ah. Guess who fits both of those categories.

These rules aren’t just an insurance policy for the company—they’ve already been used to shield the president from suspension. In September, when Trump warned in a tweet that “Little Rocket Man … won’t be around much longer,” the company said that the threatening tweets didn’t violate its guidelines because they were “newsworthy.”

Now the company has slapped on another policy, and Trump—and other government and military leaders—will get the same monopoly on violence on Twitter that they already enjoy out in the world.

At least we’ll have a thorough understanding of why the nukes are headed this way.