Comment Re:Bias? (Score 1) 99
Why not?
For all we know their detection may be bayesian based and still has "learning" to do in the field. Maybe this learning can take place in a matter of days with a sampling size as large as Android's. I think a trade-off of some start-up time in return for a system that can cope better with new attempts to circumvent its detection the better.
FWIW this article is a beat-up - Google have multiple layers to their malware detection, and they've only tested one layer.