It's the gene drive part that's controversial.
It refers to making sure all children (and grandchildren etc) inherit a trait, even though the trait isn't necessarily evolutionary beneficial for the organism. The way this is implemented in this case by having a gene that codes for proteins that insert the gene in chromosomes that do not already have it.
This means that by just releasing a few specimen with the gene drive, the entire population can be gene modified in not-that-many generations. A year may be sufficient for insects, since their generations are so short.
If the trait is negative, it will be selected out of the population. It's only if it's beneficial, or neutral, that that it will spread. What's particularly special about gene drive, is that it drives itself onto 100% of progeny, rather than the 50% expected by traditional mendellian genetics.
Only if it is extremely negative, by more than halving amount of offspring, will it be selected out.
That's in the short term. Presumably, some sort of resistance would evolve in the long term.
At the very least, this experimentation is likely to show some interesting real-world population genetics that we can measure.
In a sense the technology is like nuclear material: it can be used for ill; it can be used for good; and it can cause harm even if trying to do good. But this is also true of our more modest applications of human ingenuity and power, such as civil engineering. I am choosing the nuclear parallel because of the wide-ranging impact that facile genome editing will have - both positive and, inevitably, with some negatives.