[waserwill]

The encoding used by these sorts of projects avoids 'ATG' (start site), and possibly the Shine-Dalgarno Sequence (where ribosomes attach), if they do it right. The bacterial DNA can mutate, but the likelihood of that mutation creating a viable protein, let alone a dangerous one, is very low. Honestly, since these stored data aren't coding, the chances of them surviving for long, even in a population of bacteria, is low. Bacteria tend to favor smaller genomes, with higher percentage of coding DNA.