Using more traditional techniques, you insert a selection marker (e.g. an antibiotic resistance) which helps you selecting clones where the modification successfully took place (all the clones which do not have the antibiotic resistance will die after applying the antibiotic, leaving only clones where the modification was successful).
In a next step this marker is usually removed, but this still leaves some "scars" (e.g. so called "FRT scars", a short DNA sequence), that could theoretically be used to prove that modifications were made.
Because CRISPR-Cas9 is very efficient, these steps are not needed anymore, so in theory there are no artifacts which would hint that any "non-natural" modifications would have occurred.
That isn't really correct. If you delete a section of DNA, then the difference is that piece of DNA has been excised. Given that excision (or duplication) can occur naturally, there is no way of telling whether the change is due to CRISPR or not. It doesn't leave tell-tale signs.
If there were a strict regulatory regime, there's simply no way of enforcing compliance. Of course it's simple when you put in something that's obviously exogenous - cheap genome sequencing will pick it up, hence why that will be regulated. Anything else, and you're relying on an honesty system.
Remember that the system has been discovered approx. 10 years ago and the idea to use it for genome modifications (especially Cas9 from Streptococcus pyogenes) is only a couple of years old.
