I'm immediately curious what percentage of non-GMO certified produce are cultivars that were produced this way. This approach strikes me as being much more likely to accidentally produce unexpectedly harmful effects than modern gene recombination.
With atomic gardening, you're changing the genetic code at random, and I would guess that you really have no decent way of knowing exactly what's changed and what effects it might have. So you're only going to notice things that are really obvious, or that you specifically looked for. And you can't just do a dragnet search for every possible effect, because that would get you sent to the 4th circle of scientific hell. [1]
Personally I'm not worried about whether Rio Star grapefruit is safe to eat. (About as safe as any food can be, that is. This morning I stumbled across something about french press coffee being bad for cholesterol levels, and can't help but think that, if you look hard enough, you can probably find some way in which every food is killing you. Which isn't to say that nothing matters, but one of the things that does matter is effect size. But I digress.) It's been around for a long time and people have been eating it with no apparent ill effects. But I'm pretty sure I'm not using the same risk calculus as folks who belong to the non-GMO certification program's target market.
[1]: https://journals.sagepub.com/doi/10.1177/1745691612459519
One thing to consider though is that our bodies have evolved to eat all sorts of DNA - fragments and full genes - without directly incorporating it into our own genomes.
I'd personally be more concerned about plants expressing new allergenic proteins or that sort of thing, but it seems like there are already some decent methods in use to detect that.