Harmful and false, I assume? Surely harmful but truthful content wouldn't be censored.

Conversely, will beneficial but false content also be taken down, or will that be allowed to stand?

And we're all working with same definitions of "harmful" and "beneficial", right?

It’s really filtering for “advertiser friendly” which is a proxy for “content that doesn’t upset people”. The intersection between that and “truth” is definitely not a perfect overlap.

Harmful and "false" content shouldn't be censored by tech platforms either. There's no API for truth.

There's some role for courts, through libel/defamation lawsuits. And even then, the general criteria is false factual statements (not merely opinions) causing demonstrable harm _intentionally_.

Feels like there should be a textbook called: From detecting cats to detecting wrongthink, AI in the 201Xs