Tag · 10 stories
Every story tagged Content Moderation on AI Chat Daily.
Reddit says AI cut harmful-content exposure by 40 percent, but wrongful bans and mass deletions show the limits of automated moderation.
Tech Transparency Project found paid ads promoting nudify apps, reviewed and approved by Meta, running as recently as this week.
The LLM-based system moves beyond keyword matching and will hit all new communities before a full 2026 launch.
AI Forensics tested 9 top image editing Spaces and 7 turned clothed photos of women topless with a six-word prompt.
The tool scans posts, notes, and comments longer than 100 words and rolls out on web and iOS, with Android coming soon.
The scanner uses a Jumio identity check and a selfie to flag unauthorized deepfakes, launching first with a slice of US creators.
Instagram will keep labeling AI posts but won't let users hide them, even as detection gets harder and Muse Spark tagging raises abuse concerns.
Spreadsheets, chessboards and game textures were flagged as harmful content after a bug bypassed the human review step meant to catch false positives.
The Meta AI app's For You section generated fake articles with images of two Queen Elizabeth IIs; Meta says it will deprecate the feature.
After months of speculation that the company would let the body wither, Meta extended its lifeline with fresh capital running into 2028.
The tightly edited AI news email engineers, founders, and investors actually open. One email. Every weekday. Five minutes to finish.
The briefing read inside teams at
