It's sort of silly to say that 23andMe hasn't been "validated" to tell you things to look out for later in life. The article said that the quality of the SNP is highly accurate, with caveats about being aware of the population studied in the studies. If it tells you that you have a statistically higher chance of developing Alzheimer's (say, 2 or 3 times average), then you can prepare for that level of increased risk to the extent that it is rational to do so (maybe 3 times the risk shouldn't cause any behavior change), whether or not this has been approved by some sort of governmental or scientific agency.
Edit: To be more general, the category of "Bayesian evidence" is larger than the category of "scientific evidence." Acting based on Bayesian evidence is the Right Strategy.