Language models will hide 'bad news' in reports by default

So maybe a useful prompt for summarizing these language model generated reports included "Put the bad news first"