Automated trust-and-safety at scale: classifiers, deepfake detection and policy enforcement.
266 articles
Nairobi firm Sama abruptly sacked over 1,000 low-paid Kenyan workers after Meta ended their outsourcing deal. Termination followed allegations of staff viewing private smart glasses footage.
The Guardian Technology · 155d ago
Meta has declined to approve legal changes letting its Oversight Board work with AI companies on policy. Board eyes gen AI moderation amid new harms from LLMs and chatbots.
Engadget · 156d ago
A damning report reveals Apple and Google's app stores promote AI nudify apps via autocomplete suggestions and sponsored ads, despite explicit content bans. Some of these apps are rated suitable for kids.
Digital Trends · 156d ago
Apple threatened to remove Elon Musk's Grok AI app from its App Store in January due to inadequate moderation of nonconsensual sexual deepfakes on X. Following complaints and news coverage, Apple demanded that X and Grok teams develop a content moderation improvement plan.
The Verge · 157d ago
EU's ePrivacy exception for scanning child sexual abuse material expired on April 3, leaving a regulatory gap amid privacy debates. Google, Meta, Snap, and Microsoft call it a '失職的失敗' and pledge voluntary continuation.
IT之家 · 160d ago
Top Senate Republican launches inquiry into Meta, Amazon, xAI and five other tech firms. Focuses on reporting of suspected online child sexual exploitation cases.
Bloomberg Technology · 163d ago
Roblox launched an AI moderation system that scans entire game scenes in real time. It detects harmful content missed by older tools.
Digital Trends · 173d ago
English Wikipedia banned generative AI for writing or rewriting articles due to violations of core policies. Limited exceptions allow AI for refining personal writing or translation, but require human verification for accuracy.
Engadget · 177d ago
Wikipedia has updated its rules to explicitly ban editors from using AI to directly write or rewrite articles due to frequent policy violations. AI tools are still allowed for basic polishing of wording/grammar without adding content and for translations if the editor is proficient in the source language.
IT之家 · 177d ago
Meta is rolling out new AI systems for content enforcement to detect violations more accurately and prevent scams. These tools enable quicker responses to real-world events while reducing over-enforcement.
TechCrunch AI · 184d ago
Meta is launching new AI tools to enhance support and content enforcement across its apps. These tools aim to improve safety and user experience on platforms like Facebook, Instagram, and WhatsApp.
Meta Newsroom · 184d ago
Meta Platforms will cut back on third-party content moderators. The company plans to rely more on advanced AI systems.
Bloomberg Technology · 184d ago
OpenAI's internal mental health experts unanimously opposed launching a 'naughty' version of ChatGPT. The company draws a distinction between AI-generated 'smut' and pornography.
Ars Technica · 187d ago
Facebook is upgrading creator protection tools to combat AI-generated spam and impersonation accounts. Meta is enhancing detection and handling of fake accounts while updating guidelines to define original content clearly.
cnBeta (Full RSS) · 190d ago
YouTube is expanding its AI-generated fake video detection to select politicians and journalists in testing. The tool identifies videos featuring faces resembling theirs and enables deletion requests for guideline violations.
ITmedia AI+ (日本) · 192d ago
YouTube has introduced a new tool enabling public figures to report fake videos impersonating them. This addresses growing pressure on social media platforms to combat deepfake videos that use AI-generated deceptive images of real people.
New York Times Technology · 193d ago
YouTube is expanding its AI deepfake detection tool to politicians, government officials, and journalists. This allows them to flag unauthorized uses of their likeness in videos for removal.
TechCrunch AI · 193d ago
X will suspend paid creators from revenue sharing for posting unlabeled AI-generated armed conflict videos, with 90-day bans for first offenses. Policy uses Community Notes and AI metadata detection, framed for 'times of war'.
Engadget · 200d ago
OpenAI informed Canadian lawmakers it would have referred a banned ChatGPT user—who became the main suspect in a major mass shooting—to police under new policies. This highlights updated safety measures for risky users.
Bloomberg Technology · 205d ago
Starting next week, Instagram will notify parents if adolescent users repeatedly search self-harm or suicide-related keywords in a short time. Meta plans to extend this alert mechanism to its AI chatbot products later this year.
cnBeta (Full RSS) · 205d ago
Pixiv will revise its guidelines on March 18 to prohibit mass posting of works and attaching false statuses like incorrect AI generation labels. Violating works may be hidden from view.
ITmedia AI+ (日本) · 213d ago
Apple and Google missed a UK deadline to block child nudity on their phones. The UK government plans legislation imposing fines and potential criminal liability, with similar measures possibly extending to Snapchat and Instagram.
Wired · 11d ago
Islamic terror supporters are reportedly using generative AI tools to produce and distribute extremist content. The campaign, described as “Slop Jihad,” is targeting new audiences on TikTok.
Wired · 11d ago
ASCII smuggling, an invisible Unicode technique once associated with attacks against AI systems, is increasingly being used by spammers. The trend highlights how hidden text can bypass human review and potentially complicate automated content filtering.
Ars Technica AI · 15d ago
Instagram will demote accounts presenting AI-generated influencers if they do not clearly disclose that status. The platform is not applying the same crackdown to other forms of AI-generated low-quality content.
Engadget · 19d ago
Instagram is renaming its “AI creator” label to “AI-generated profile” to clarify when an account depicts a non-human, AI-generated person. The platform will also target accounts that fail to disclose their AI-generated profiles and may limit their reach or features.
The Verge · 19d ago
AI-generated dating videos are misleading older adults into believing that real people are seeking partners, exposing them to ridicule and harassment. The article argues that older adults’ emotional and social needs deserve respect, while platforms and users should address AI deception without discriminating by age.
虎嗅 · 22d ago
Meta says approximately $5.3 billion of its teen safety settlement will be withheld unless TikTok and YouTube adopt comparable rules and make matching payments. The structure follows Meta's agreement to settle a case brought by U.S.
The Next Web (TNW) · 24d ago
美國勞工統計局預測，2025至2035年總就業將增加約590萬個職位，許多高AI暴露職業仍會成長。AI更可能先拆解傳統崗位中的任務，重新分配人類與機器的工作邊界。
虎嗅 · 10d ago
China’s AI-generated short-drama market is moving from mass production toward stricter copyright, originality, character-similarity, and content-quality controls. Platforms are reducing promotion for low-quality works, while creators with original stories, distinctive visual identities, and greater production investment are more likely to survive.
虎嗅 · 25d ago