It’s a matter of cost. To usefully redact people’s names from these records, and only the people who are innocent or unrelated to the case, requires advanced natural language processing. That’s currently done by human beings in a few passes. Because of the sheer number of documents, the process is slow.