The world is nuanced.

Imagine no e2e for a moment for FB. Policy can be smart enough to pick up that this communication is not represntative or normal. That's part of detection.

Second, a single message to someone on a random phone is not going to flag anything.

Third - there is no assumption of guilt. Not even an arrest is assumption of guilt.

Finally - those are extraordinary corner cases. They will happen, but the get resolved the moment the guy says 'oh, I found this phone' - because that will be 100% clear in that context.

Obviously - things can go awry. Meta flag something as bad, sends it to police - they do not follow procedure, or don't apply something correclty and arrest a guy at his place of work. But in the scenario you described, its literally not a problem - there are 'common sense checks' through the whole thing. The algo, the human making the notification to the police, the police, the judge if a warrant is required. People are not going to be arrested because they found a phone and texted their niece - if that happens, then we have another set of problems.

We can 100% have our 'friendly community' with Facebook.

Now - with an e2e thing like Signal, well, yes, it could theoretically be a problem, but the likelihood of some rando finding a phone, that's not locked, and being able to text some other 12 year old, an effectively 'pose' as their 'contact' - well that's a rare case scenario.