It’s a risk management strategy where you only do checks afterwards.

“Trust” means that you don’t make processes wait on passing checks before proceeding, because that would be expensive and/or slow.

“Verify” means that you have a separate process that comes through and runs checks afterwards, maybe on only some of the things you trusted, to catch issues.

It’s ideal when you have high-volume and/or low-latency processes where failures are low stakes but you still want to catch systemic issues eventually.

It’s related to the idea that “the optimal amount of fraud is non-zero”.