Do we gain more by having more people understand software? Do some risks get mitigated? I'd argue yes on both counts. And that's why I encourage people to get their kids to poke around with Scratch and maybe some Python. It's why smart people doing liberal arts degrees at Harvard still want to go and do CS50.
Most people don't want to learn to code because it simply doesn't interest them as a subject, a perfectly valid way of living your life.
So, no, not everyone needs to be a car mechanic. But everyone who wants to drive a car needs to learn how to drive it safely, and would benefit from checking the oil, tire pressures, and knowing something about how their car worked to keep it in good, safe working order. If the only people who know all those things are professional mechanics, we're in a worse place.
Most professionals work with data but have low technical literacy. Imagine if most professionals could just query databases themselves. In a generation or two it could be made possible.
However "learning to program" and "understanding software" is not the same thing. Learning to program helps one to better understand software, but it's not a necessity or requirement of it.
"Everyone needs to learn" vs "People who needs to interact with software needs to better understand it" are very different points on the "understanding software" axis.