The problem isn't the technology, it's the irreversible ripple effect. A single edit to an embryo changes not just that child, but every generation after them. We don't fully understand gene interactions or off-target mutations, and we can't test for long-term consequences before birth. That's like deploying code you've never run in production, except the users are real people who never consented to the change.
Many genetic diseases can be addressed through PGD or gene therapy in existing people. Editing embryos doesn't fix a patient; it designs a person. That crosses a line from treatment to engineering. And once we start selecting for disease, we inevitably start selecting for traits. No ethical framework handles that slope well.
Until we can guarantee safety across generations, and secure genuine consent from those affected, this should stay off the table. The burden of proof is on the edit, not the status quo.