At a previous company we used https://postgresql-anonymizer.readthedocs.io/en/latest/ https://postgresql-anonymizer.readthedocs.io/en/latest/, it worked out wel

At a previous company we used https://postgresql-anonymizer.readthedocs.io/en/latest/ https://postgresql-anonymizer.readthedocs.io/en/latest/, it worked out well and was pretty powerful, it could even add randomized fuzziness to coordinates, swap random cities, etc. We ran it weekly to dump production down to a “scrubbed” copy which we used for our test/qa environments and then further trimmed some large tables to create a much smaller development version. It really helps to have the “shape” of production data and we saw surprises/bugs on prod deploys due to unexpected data drop to almost nothing. It was much easier to maintain than seeds imo, though it is a cat and mouse game, if a new column with sensitive data is added, you need to ensure it’s covered with a masking query too, but we solved that by auto tagging the security ops folks whenever our schema file changed on GitHub. It wasn’t perfect, but it was 95% what we needed