Here's a probably equally badly made statistic. Tesla estimates their cars had driven 130 million autopilot miles before the fatality. Some googling suggests the first non-autopilot Tesla S fatality was in july 2014 and this graph ( http://insideevs.com/wp-content/uploads/2015/06/Tesla-Model-... ) suggests that was around 400 million miles driven. So the autopilot is 3.33 times 'worse'?

Who knows, but if Tesla feels they have a moral duty to leave on autopilot if it is safer, they equally have one if it is more dangerous. It seems they are pretty happy to be in dark about which it is while they test their system.