[alexbecker]

The "individual-level information" limitation is a huge weasel. 23andMe can and does share "anonymized" aggregations of its clients' genetic information [0]. Anonymization is not a property of a dataset though; it's a property of a dataset and the state of the world, and even if (and this is a big if) the dataset is truly anonymized right now, it won't always be.

[0] "23andMe says that it is also able to share anonymous and pooled data about their self-reported health traits without asking." - https://www.forbes.com/sites/matthewherper/2015/01/06/surpri...

[marchenko]

Well said. Genetic data is by nature personally identifiable, and genomic disaggregation techniques can be expected to improve. Data troves like 23 and Me are an attractive target for DNA dragnets - at present, their SNP data is not CODIS-compatible (although it is theoretically possible that the SNAP data could be queried against a physical sample assayed for the same SNPs), but the physical samples are very valuable and customers should inspect their sample retention terms very closely.