I am trying to read this as charitably as possible, if only to clarify my own concerns. There are definitely ways it could be worse. The "video selfie" component of the assurance process is analyzed on-device. But your ID is still sent to a third-party vendor. The claim is that the data is deleted "quickly—in most cases, immediately after age confirmation." But Discord doesn't actually control that; the unnamed third party does.
A spokesperson for Discord has said of the age inference model:
Discord’s age inference model uses account information such as account tenure, device and activity data, and aggregated, high-level patterns across Discord communities. Discord does not use private messages or any message content in this process.
https://www.theverge.com/tech/875309/discord-age-verification-global-roll-out
Account tenure could mean "Are you necessarily an adult assuming you created your account when you turned 13?" As far as device and activity data, that as vague as hell! And without reviewing messages, I don't love the idea that whatever data they're collecting could be used for verification. Whether it works or not, pretty creepy.
And I guess that's where I'm landing here. I have a very public record of being too credulous with these kinds of things in the past. I'm tired of getting burned. I'm tired of my community getting burned. This won't be the last creepy thing Discord pulls out on its road to IPO. I think I'd rather take the next offramp.