Another good example of an awful implementation can be seen in Google Workspace. Sometimes the crappy ML algorithms will generate a false positive and suspend an account for a suspicious login. Fine. I get it. It’s hard to do at scale and nothing’s perfect.

The thing I’d like to have Google explain to me is why they think it’s a good idea to bounce incoming mail for a user that’s been auto-suspended by an algorithm. In what scenario would I want that, especially when the account is locked at 3:00 AM local time?

That’s something that actually happened to me this weekend. For anyone at Google, NO ONE wants their incoming email bounced because of your crappy ML algorithms.