AI News AI资讯 10h ago Updated 3h ago 更新于 3小时前 35

Meta Sued Over Training Data for Its AI and Face-Recognition Systems Meta Sued Over Training Data for Its AI and Face-Recognition Systems

Meta faces a class action lawsuit alleging it violated Illinois and California privacy laws by extracting biometric data from users' photos without notice or consent The complaint centers on the "NameTag" feature, code for which was secretly embedded in the Meta glasses AI companion app downloaded over 50 million times, designed to convert faces into biometric signatures compared against a faceprint database Meta's generative AI systems (Emu and Muse Image) are also targeted for allegedly harves Meta面临集体诉讼,被指控在Illinois和California违反隐私法,从用户照片中提取生物识别信息而未获通知或同意 WIRED报道发现NameTag代码被秘密嵌入Meta眼镜AI伴侣应用(下载量超5000万次),该系统可将面部转化为生物特征签名并与用户手机数据库中的"faceprints"比对 诉讼还针对Meta的图像生成系统(Emu和Muse Image),指控其训练过程中非法收集了图像中出现的人的生物识别信息 原告寻求每人5000美元赔偿(Illinois生物识别信息隐私法),全国集体可能达数百万人 这是Meta继2020年6.5亿美元和解和2024年14亿美元德克萨斯州和解后的

50
Hot 热度
50
Quality 质量
50
Impact 影响力

Analysis 深度分析

TL;DR

  • Meta faces a class action lawsuit alleging it violated Illinois and California privacy laws by extracting biometric data from users' photos without notice or consent
  • The complaint centers on the "NameTag" feature, code for which was secretly embedded in the Meta glasses AI companion app downloaded over 50 million times, designed to convert faces into biometric signatures compared against a faceprint database
  • Meta's generative AI systems (Emu and Muse Image) are also targeted for allegedly harvesting biometric information from Facebook and Instagram images used in training
  • This follows Meta's history of biometric data violations, including a $650 million settlement in 2020 and a $1.4 billion payment to Texas in 2024
  • The proposed national class could number in the millions, with plaintiffs seeking $5,000 per intentional violation under Illinois' Biometric Information Privacy Act

Why It Matters

This lawsuit represents a significant escalation in regulatory and legal scrutiny of how AI companies collect and use biometric data, particularly in the context of generative AI training pipelines. For AI practitioners, it underscores the growing legal risk of using social media images for model training without explicit consent, especially when biometric identifiers are involved. The case could set important precedents for how biometric privacy laws apply to emerging AI technologies and shape industry practices around data sourcing.

Technical Details

  • NameTag System: Code was embedded in the Meta glasses AI companion app (50M+ downloads) designed to capture faces via glasses cameras, convert them into biometric signatures ("faceprints"), and compare them against a local database on the user's phone that could receive updates from Meta
  • Faceprint Database: The system was configured to store faceprints and receive updates from Meta, though Meta claims no central face database exists; the origin of the faceprint data is alleged to come from Facebook and Instagram images
  • Emu and Muse Image: Meta's image-generation models trained on large quantities of Facebook and Instagram images and text; Muse Image previously allowed users to generate images based on public Instagram accounts before the feature was removed
  • Legal Framework: Illinois Biometric Information Privacy Act (BIPA) allows $5,000 per intentional/reckless violation and $1,000 per negligent violation; California privacy laws also apply
  • Timeline: Class period begins September 4, 2021; Meta removed NameTag code on June 5, 2026, one day after WIRED's report

Industry Insight

  • AI companies should urgently audit their training data pipelines for biometric information and ensure compliance with BIPA and similar state privacy laws, particularly when sourcing from social media platforms
  • The "we didn't ship it" defense may not hold in court—embedding functional biometric recognition code in a widely downloaded app, even if disabled, appears to have triggered legal liability
  • Meta's pattern of biometric violations (2020 settlement, 2024 Texas resolution, now this lawsuit) signals that regulators and plaintiffs' attorneys are treating biometric privacy as a persistent compliance failure, not an isolated incident—proactive transparency and opt-in consent mechanisms will be essential going forward

TL;DR

  • Meta面临集体诉讼,被指控在Illinois和California违反隐私法,从用户照片中提取生物识别信息而未获通知或同意
  • WIRED报道发现NameTag代码被秘密嵌入Meta眼镜AI伴侣应用(下载量超5000万次),该系统可将面部转化为生物特征签名并与用户手机数据库中的"faceprints"比对
  • 诉讼还针对Meta的图像生成系统(Emu和Muse Image),指控其训练过程中非法收集了图像中出现的人的生物识别信息
  • 原告寻求每人5000美元赔偿(Illinois生物识别信息隐私法),全国集体可能达数百万人
  • 这是Meta继2020年6.5亿美元和解和2024年14亿美元德克萨斯州和解后的又一次生物识别数据相关诉讼

为什么值得看

本文揭示了Meta在AI产品开发中可能存在的隐私合规风险,特别是面部识别和图像生成技术对生物识别数据的潜在滥用。对于AI从业者和企业而言,这凸显了在训练数据和功能开发中遵守各州隐私法规的重要性,以及透明沟通对用户信任的关键作用。

技术解析

  • NameTag系统架构:代码被嵌入Meta眼镜AI伴侣应用,设计用于将眼镜捕获的面部转化为生物特征签名,并与用户手机本地数据库中的"faceprints"进行比对,该数据库可从Meta接收更新
  • 生物识别数据来源:诉讼指控"faceprints"可能源自Facebook和Instagram图像,引用Meta员工声称NameTag可通过Meta社交关系或公开Instagram账户识别人,以及Meta专利描述的面部匹配技术
  • 图像生成系统训练数据:Emu系统使用大量Facebook和Instagram图像和文本进行训练,首席产品官Chris Cox称这些平台为AI系统的"数据优势";Muse Image曾允许用户基于他人公开Instagram账户生成图像
  • 法律赔偿标准:Illinois《生物识别信息隐私法》规定故意或鲁莽违规每人5000美元,疏忽违规每人1000美元;California提出额外赔偿和其他救济

行业启示

  • 隐私合规需前置:AI产品开发应在设计阶段就考虑各州隐私法规(如Illinois BIPA、California隐私法),避免功能开发后再面临法律风险
  • 透明沟通至关重要:Meta对NameTag的模糊表述("未向消费者推出"vs"技术上已可用")引发公众和监管质疑,企业应在AI功能开发中保持透明度
  • 训练数据治理需加强:图像生成AI使用社交媒体数据训练时,需明确生物识别信息的收集和使用边界,建立合规的数据治理框架

Disclaimer: The above content is generated by AI and is for reference only. 免责声明:以上内容由 AI 生成,仅供参考。