[nuc1e0n]

CSV isn't poorly specified, but it is mostly poorly implemented relative to the specifications that exist (Microsoft's products are some of the worst offenders for this). Data loss is not inherent in a text-based format, any more than it is in binary formats. The only difference being is that it's more difficult to produce or extract data from binary files. That's detrimental to information transfers and is why I would recommend specifically not to use binary formats.

JSON is much easier to validate and has similar ease of data transfer as with CSV. It can have minimal overhead as well if the data is stored as arrays (or an array of arrays).