There should be a difference, but there rarely is.

There is broad search going on for potential mechanisms with which such modifications could cause harm. None have been found. At some point it is time for limited experiments in the wild. We now have evidence that the "fail-safes" don't work as intended. Maybe you need to have multiple such genes.

It's a different situation with nuclear energy, with multiple theoretically impossible accidents. You could categorize those as unknown unknowns, but, I mean, they have been known to happen...