If we really want to fix this issue, we should let the government scan all our devices and communications on a constant basis.

You are probably joking, but for argument sake.

And then? Generate 100000 false positives per day (which is what will happen); arrest them all until proven innocent? Investigate them all until proven guilty? What? 100000 flagged data streams (videos, documents, sounds, images etc) is serious work to comb through as an AI is just not good enough to ruin someone's life while innocent. This will also happen with the current proposal anyway if executed broad enough.

There is nothing remotely specific about "manipulative behaviors".

That's why it's followed by specifics (...that the abuser uses to gain access to a potential victim, coerce them to agree to the abuse, and reduce the risk of being caught.)

Of course taking two words out of the statement without context makes it less specific.