>As mentioned, detecting ‘grooming’ would have a positive impact on the fundamental rights of potential victims especially by contributing to the prevention of abuse; if swift action is taken, it may even prevent a child from suffering harm. At the same time, the detection process is generally speaking the most intrusive one for users (compared to the detection of the dissemination of known and new child sexual abuse material), since it requires automatically scanning through texts in interpersonal communications. It is important to bear in mind in this regard that such scanning is often the only possible way to detect it and that the technology used does not ‘understand’ the content of the communications but rather looks for known, pre-identified patterns that indicate potential grooming. Detection technologies have also already acquired a high degree of accuracy, although human oversight and review remain necessary, and indicators of ‘grooming’ are becoming ever more reliable with time, as the algorithms learn.
That whole paragraph sounds a lot more terrifying and intrusive. At least they admit the flaws, but I haven't quite had the best experience with the EU's way of handling this stuff.