First, being in the field and a part of the teams that first published the general human genome and the individual human genome, I know that our current knowledgebase of genomic variant <-> disease correlation is not very vast. We don't even know how much more or less important are the particular type of variants "SNPs" that personal genomics companies mostly mine, in comparison with other types (Insertion/deletions , structural variations, copy number). At more serious genomics research institutes, we are still trying to figure out the genetic pathways that might be affected by this mutation or that - far from actually predicting their effects on actual physical condition. So does that mean personal genomic companies are far ahead of institutes in their research? I am not so sure.
Second, I think we need regulation here for the same reasons that we need net neutrality. I wouldn't be surprised if with time, the reports start advising you on which of the alternative drugs would be most beneficial for your genotype and disease. I would still be okay with it if companies like 23andme published their research or atleast cited publications - but I am not sure they believe in public access when it comes to their findings. (I can be wrong on that - so correct me if you know better)
Finally, I actually like the prospect of public driven science in addition to just organization/scientist driven, as that is likely to fund some actual science in addtion to pseudo science. So one solution I can think of is for FDA to approve a most trusted mutation screen for all these personal genomic companies to abide by. Let them focus on their core competency which is sequencing and customer interface and not research, If however they come across findings that might be beneficial, let them publish it in a peer-reviewed journal and apply to FDA to add it to the universal mutation screen.