Tag: #moderation10 results
Tech Transparency Project reports Apple App Store and Google Play still host and promote 38 nudify apps using AI for deepfakes, despite bans. These apps, some rated 'E' for Everyone, generated $122M revenue and 483M downloads. Apple removed 15 apps, Google suspended some after the report.
Meta's AI smart glasses users in Europe are reportedly sharing intimate videos and sensitive financial data with human moderators in Kenya. This occurs because activating AI features requires agreeing to terms allowing human review of captured data for LLM training via annotation. The practice raises GDPR compliance concerns due to lack of transparency.
A technical glitch within Discord's automated safety systems has resulted in the incorrect banning of approximately 8,200 user accounts since May. The company is currently addressing the issue to rectify these automated enforcement errors.
During a criminal trial, it was revealed that a suspect attempted to use ChatGPT to generate violent imagery of a burning city. OpenAI's safety guardrails successfully blocked these requests, demonstrating the effectiveness of current content moderation policies.
Meta is rolling out its '13+' content settings globally across Instagram, Facebook, and Messenger. This system, which functions similarly to a movie rating, aims to restrict sensitive content for teenage users by default.
A surge in AI-generated child sex abuse imagery is emboldening predators. Law enforcement struggles to distinguish fakes from real cases involving children in danger. Investigators can't keep up with the volume.
Meta's advisers criticize its current methods for policing AI-generated videos as inadequate. They highlight vulnerabilities especially during crises. The recommendation calls for enhanced oversight to combat fake content.
Researchers are reporting significantly longer wait times for arXiv submissions to move from 'on-hold' to 'submitted'. This delay is speculated to be caused by the recent influx of low-effort AI-generated papers.
US judge dismissed X Corp.'s antitrust lawsuit against advertisers in GARM for boycotting due to moderation issues. The case was tossed with prejudice, barring refiling or appeal. X, under xAI, faces separate Grok AI safety concerns over explicit imagery.
Instagram will discontinue end-to-end encryption for direct messages starting May 8, 2026. Meta made the quiet announcement amid years of criticism from child safety groups and law enforcement. The change updates a 2022 news post on its help page.