Meta Claims 90%+ AI Detection of Harmful Content—But the Math Has a Catch

Meta's self-reported proactive rates look impressive, but critics note the denominator is disputed, non-English content lags badly, and independent verification remains scarce.

ndtv-profit-business · 7/5/2026, 11:31:31 AM · Ask about this →