Similarity-based data loss prevention

A technique performs similarity-based data loss prevention on content from a content source. The technique involves generating multiple variants from the content, the multiple variants including a set of variants for each parsed word of the content, each variant of that set (i) including multiple characters and (ii) differing from other variants of that set by at least one character. The technique further involves performing evaluation operations to determine whether any of the variants includes sensitive data. The technique further involves performing, in response to the evaluation operations, a control operation which (i) releases all of the parsed words of the content to a destination when none of the variants is determined to include sensitive data, and (ii) blocks at least one parsed word of the content from reaching the destination when at least one variant is determined to include sensitive data.