This is misleading, yet everyone seems to repeat it. Discord's implementation of ID verification did not retain IDs. Reporting on this was so poor, but what appears to have happened was that people that failed age estimation / ID checks had to raise a support ticket and get manually reviewed. That support platform was pwned and the active support tickets were leaked. Who knows how long these support tickets were set to live for, but up to 70,000 active tickets getting leaked feels like a drop in the bucket. It's also not immediately clear to me what the alternative is (other than not getting hacked), when you require human intervention to review problematic IDs. Even if the ID only lived on their server for 24 hours during manual review, across a userbase of >200 million users, that's a lot of IDs at risk at any given moment, especially during these initial roll outs of age verification.