There are different narratives. What I believe I heard on BBC radio was that the sender intended to just send a few records to the recipient but accidentally sent the full dataset of 18,000 (or 19,000) records.

Other sources, like the one I linked, say the error was that the wrong recipient was reached.

It’s surprising that as big as the story is, reporting of the tech details is lousy. A couple sources say the data was a spreadsheet, which seems to support what the Guardian says.