AI News AI资讯 1d ago Updated 1d ago 更新于 1天前 42

Dutch data protection authority advises Twitch users to opt out from Amazon AI 荷兰数据保护机构建议Twitch用户退出与亚马逊AI的数据共享

Amazon is using Twitch users' personal data (streams, images, chats) to train AI models, with the opt-out setting enabled by default The Dutch data protection authority (AP) strongly urges users to disable the "Training for Generative AI" setting due to irreversible data exposure risks Facial images and voice recordings constitute sensitive personal data that cannot be removed once stored in AI systems Twitch admitted the default-enabled setting exists because "otherwise nobody would give their Amazon旗下Twitch平台默认将用户直播数据(面部图像、语音、聊天内容)用于训练生成式AI模型 荷兰数据保护局(AP)强烈建议用户主动关闭"Training for Generative AI"选项以保护隐私 一旦数据进入Amazon AI系统即无法删除,用户永久丧失数据控制权 Twitch承认默认开启该设置的原因是"否则不会有人同意",暴露知情同意机制的失效 面部图像属于敏感个人数据,存在被AI模型泄露给网络诈骗者的风险

65
Hot 热度
60
Quality 质量
55
Impact 影响力

Analysis 深度分析

TL;DR

  • Amazon is using Twitch users' personal data (streams, images, chats) to train AI models, with the opt-out setting enabled by default
  • The Dutch data protection authority (AP) strongly urges users to disable the "Training for Generative AI" setting due to irreversible data exposure risks
  • Facial images and voice recordings constitute sensitive personal data that cannot be removed once stored in AI systems
  • Twitch admitted the default-enabled setting exists because "otherwise nobody would give their consent," raising transparency concerns
  • AI models trained on this data could potentially leak personal information to malicious actors like online scammers

Why It Matters

This case highlights the growing tension between AI development and data privacy, as tech companies increasingly rely on user-generated content for model training without meaningful consent. It serves as a cautionary example for AI practitioners and companies about the regulatory and ethical risks of opt-out-by-default data collection practices, particularly when sensitive biometric data is involved.

Technical Details

  • Twitch's "Training for Generative AI" setting is enabled by default, automatically enrolling all users whose data is used to train Amazon's AI models unless they manually opt out
  • Data collected includes live streams (facial images, voice, name), private space imagery (e.g., bedrooms), and user chats/text messages
  • Once submitted to Amazon's AI systems, the data cannot be deleted, creating permanent and irreversible exposure
  • The setting is located under Settings → Security and Privacy → "Training for Generative AI"
  • Twitch's own statement acknowledged the default-on design was intentional to maximize data collection, as voluntary consent rates would be significantly lower

Industry Insight

  • Companies should anticipate stricter regulatory scrutiny around default-opt-in consent models for AI training data, especially when biometric or sensitive personal data is involved
  • The irreversible nature of data ingestion into AI models demands greater transparency and user control mechanisms to mitigate reputational and legal risk
  • This case may accelerate similar regulatory actions globally, making proactive privacy-by-design approaches a competitive advantage rather than a compliance burden

TL;DR

  • Amazon旗下Twitch平台默认将用户直播数据(面部图像、语音、聊天内容)用于训练生成式AI模型
  • 荷兰数据保护局(AP)强烈建议用户主动关闭"Training for Generative AI"选项以保护隐私
  • 一旦数据进入Amazon AI系统即无法删除,用户永久丧失数据控制权
  • Twitch承认默认开启该设置的原因是"否则不会有人同意",暴露知情同意机制的失效
  • 面部图像属于敏感个人数据,存在被AI模型泄露给网络诈骗者的风险

为什么值得看

这篇文章揭示了大型科技公司利用用户数据训练AI的普遍做法及其隐私风险,对关注数据保护、AI伦理的从业者和用户具有重要警示意义。

技术解析

  • 数据范围:Twitch直播中的面部图像、语音、姓名、私人空间画面、聊天文本等,其中面部图像属于敏感个人数据
  • 默认开启机制:"Training for Generative AI"设置默认启用,用户需主动前往Settings > Security and Privacy手动关闭
  • 数据不可逆性:一旦数据进入Amazon AI训练系统,用户无法删除或撤回,失去对数据的控制权
  • 风险场景:AI模型可能泄露用户数据给网络诈骗者,造成身份盗用等二次伤害

行业启示

  • 企业应在产品设计中采用隐私优先原则,避免默认开启数据收集功能,而非依赖"用户不会同意"作为默认开启的理由
  • 监管机构对AI数据使用的审查趋严,荷兰AP的介入表明各国数据保护机构正在加强对AI训练数据合法性的监督
  • 用户数据主权意识觉醒,平台需在商业利益与用户信任之间找到平衡,否则将面临监管处罚和用户流失的双重风险

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。

LLM 大模型 Training 训练 Policy 政策 Regulation 监管 Ethics 伦理