> Instead, the default should be, that if there is no header or it cannot be parsed, then the content is unsafe.
That's something the client should be doing. You can configure your own device (or your kid's) to have whatever default you want.
The actual problem is that classifying the entire firehose of user-generated content is precarious and uneconomical, so the default tag is going to be whatever minimizes liability. If untagged implies "unsafe" and "unsafe" minimizes liability then the majority of content will be untagged. If untagged implies "safe" then the majority of content will end up explicitly tagged as unsafe, because tagging it as unsafe minimizes liability.
But either way if you disable "unsafe content" you'll end up disabling almost everything, specifically including the huge amount of "safe" content which isn't tagged as safe because accurately classifying it is uneconomical.
That's something the client should be doing. You can configure your own device (or your kid's) to have whatever default you want.
The actual problem is that classifying the entire firehose of user-generated content is precarious and uneconomical, so the default tag is going to be whatever minimizes liability. If untagged implies "unsafe" and "unsafe" minimizes liability then the majority of content will be untagged. If untagged implies "safe" then the majority of content will end up explicitly tagged as unsafe, because tagging it as unsafe minimizes liability.
But either way if you disable "unsafe content" you'll end up disabling almost everything, specifically including the huge amount of "safe" content which isn't tagged as safe because accurately classifying it is uneconomical.