To add context, this refers to a lesson from WW2 by Abraham Wald.

American bombers would come back damaged, so Wald calculated the average damage (bullet holes down on the planes) on the returning planes to help offer clues as to where to focus the armor on the planes (since armor can only be provided in limited quantities due to weight). The thing he realized was that, the damage should be completely random and evenly distributed on the planes. They weren't. The planes returning had less damage on their engines, and mid body(image on the flag for reference) for instance. His insight was, don't apply armor where the returning planes are mainly damaged, as the planes can obviously survive that damage, instead apply the armor where the surviving planes were less damaged, like the engines.

This is called the survivorship bias, basing information on only those who are here and healthy doesn't mean the current system works: we're ignoring the result from those that the system completely failed, because the others didn't survive.