This embodies an inherent contradiction.

People use it seeking to save time, money and effort.

Verifying everything detracts from any savings.

Human nature being what it is, people are motivated to short circuit or just skip the verification part.

Bottom line: AI has the potential to create as many problems as it solves.

Example for software devs: AI is just so-so at generating code but it is absolutely terrible at maintaining it. "Maintenance" is not easily represented in model training. So what it really does is create technical debt at a faster pace.