@Darzil
That is the way it works currently - they have to default to anything they don't have an autosoft for, which I'm not changing, I'm just saying it is silly that a pilot program, whose main purpose is to maneuver the vehicle it occupies safely, should be defaulting when doing that task. Defaulting means that your average drone has a dice pool of 2 to avoid crashing into things. Many drones today have better collision avoidance than that - in fact, I'd go so far as to say that any self-maneuvering drone today has better collision avoidance than that. It just makes sense that by default, drones would be better at such common activities. For example, your default vehicle has a pilot of 3. That means it has only a 66% chance of avoiding a pedestrian that steps into the road unexpectedly. Or, more concerning, it has a significant chance of crashing when weather conditions get foul (I'd think that an ice patch is the sort of thing that would be considered "out of the ordinary" enough to call for a dice roll). Point is, in order for the world of the 2070s to be as dependent on pilots as it is, they should be significantly more robust, otherwise nobody would bother.
Pg 199 of core puts an Average task at a threshold of 2, including things like avoiding an obstacle. A dice pool of 6 succeeds at this about 50% of the time. Now, what is average? This is the stunt table, so this is an average "stunt", and that obstacle is obviously not you average stationary object, so I don't expect drones to face a situation like this often, but I do expect that they would face Easy stunts ('merging, passing, sudden stops," etc.) on a daily or weekly basis. They should have a decent chance at performing these things successfully. Drifting is listed at this level, but I could also see that as applying to hydroplaning in wet weather. A dice pool of 6 only has a 91% chance of success on a threshold of 1, so even granting a pilot 3 (the most common pilot rating) a free maneuvering 3 autosoft means that during rain, you'll still have a ton of crashes caused by crappy pilots, but at least it won't be an absolute cluster. Considering it highly specialized (for +2, per my other suggested rule change) bumps that to a 96% chance of avoiding an accident - still practically guaranteed to happen frequently in a city of millions, but at least at a point where you can write it off as a limitation of the game system (you can't have vanishingly small chances of failure unless you have massive dice pools) rather than an unconscionable problem with the technology.