AI News AI资讯 4d ago Updated 3d ago 更新于 3天前 61

If you use Google, you’re training its AI. Here’s how to opt out. 如果你使用Google,你正在训练它的AI。以下是如何退出的方法。

Google has updated its privacy settings to allow the storage of user-uploaded media (images, audio, video) for AI model training, affecting services like Search, Lens, and Translate. The change separates "Search Services History" from general "Web & App Activity," meaning previous opt-outs for general activity no longer prevent media collection for AI purposes. Users are effectively opted-in by default to this expanded data usage, though they can manually disable "Save Media" or set automatic de Google更新隐私设置,默认将用户上传至搜索服务(如图片、音频、视频)的数据用于训练AI模型。 该变更通过“搜索服务历史”和“个性化推荐”两项新设置实施,涵盖Google Lens、语音搜索及翻译等场景。 用户需手动取消勾选“保存媒体”选项方可退出数据用于AI训练,旧版“网页和应用活动”设置不再自动涵盖此权限。 此举反映了科技巨头从网络抓取转向收集用户生成内容(UGC)以优化AI的行业普遍趋势。

72
Hot 热度
68
Quality 质量
75
Impact 影响力

Analysis 深度分析

TL;DR

  • Google has updated its privacy settings to allow the storage of user-uploaded media (images, audio, video) for AI model training, affecting services like Search, Lens, and Translate.
  • The change separates "Search Services History" from general "Web & App Activity," meaning previous opt-outs for general activity no longer prevent media collection for AI purposes.
  • Users are effectively opted-in by default to this expanded data usage, though they can manually disable "Save Media" or set automatic deletion intervals (3, 18, or 36 months).
  • This reflects a broader industry trend where tech giants like Meta are increasingly leveraging user-generated content and uploaded media to train generative AI models.

Why It Matters

This update significantly expands the scope of data collection for AI development, moving beyond text-based search queries to include rich multimedia assets created by users. For AI practitioners and researchers, it highlights the critical importance of understanding data provenance and the ethical implications of using user-generated content for training. It also serves as a warning for developers and product managers to ensure transparency in privacy settings, as opaque changes can lead to significant user trust issues and regulatory scrutiny.

Technical Details

  • Data Scope Expansion: The update explicitly includes "images, files, and audio and video recordings" from Google Lens, Search Live (voice), and Google Translate into the pool of data used to develop and improve AI models.
  • Settings Architecture: Google has decoupled "Search Services History" from "Web & App Activity." Previously, controlling one affected the other; now, "Search Services History" is a distinct, enabled-by-default setting that governs media retention for AI training.
  • Retention Mechanisms: Users can configure automatic data deletion cycles (3, 18, or 36 months) and independently toggle the "Save Media" checkbox, allowing for granular control over how long specific media types are retained before deletion or potential use in training.
  • Service Integration: The policy applies across multiple search-centric services including Maps, Shopping, Flights, Hotels, Translate, and News, indicating a systemic backend integration rather than a feature-specific addition.

Industry Insight

  • Privacy by Design vs. Dark Patterns: Companies must prioritize clear, upfront consent mechanisms for data usage in AI training. Bundling AI training permissions with general service functionality without prominent disclosure risks eroding user trust and inviting regulatory action.
  • Data Strategy Shift: The industry is shifting from passive web scraping to active collection of high-quality, user-generated multimodal data. Organizations should anticipate increased competition for user attention regarding data privacy controls and consider offering value propositions for data sharing.
  • Compliance Preparedness: With regulations like GDPR and CCPA evolving to address AI-specific data concerns, firms should audit their data retention policies and ensure that opt-out mechanisms are robust, easily accessible, and clearly communicated to avoid legal liabilities.

TL;DR

  • Google更新隐私设置,默认将用户上传至搜索服务(如图片、音频、视频)的数据用于训练AI模型。
  • 该变更通过“搜索服务历史”和“个性化推荐”两项新设置实施,涵盖Google Lens、语音搜索及翻译等场景。
  • 用户需手动取消勾选“保存媒体”选项方可退出数据用于AI训练,旧版“网页和应用活动”设置不再自动涵盖此权限。
  • 此举反映了科技巨头从网络抓取转向收集用户生成内容(UGC)以优化AI的行业普遍趋势。

为什么值得看

对于AI从业者和普通用户而言,这揭示了大型科技公司获取高质量训练数据的最新策略转变,即深度挖掘用户交互产生的多媒体数据。同时,它警示了隐私设置的复杂性增加,用户需重新审视并主动管理自己的数据授权状态,以避免无意中成为AI模型的训练素材。

技术解析

  • 数据范围扩展:除了传统的文本搜索历史,新增收集的数据类型包括通过Google Lens拍摄的图片、使用Search Live进行的语音录音、以及Google Translate中的语音练习音频。
  • 架构调整:原有的单一“网页和应用活动”设置被拆分为独立的“搜索服务历史”和“个性化推荐”两个模块,其中“搜索服务历史”默认开启且包含媒体保存选项。
  • 存储与保留机制:保存的媒体数据不仅用于即时功能支持,还明确用于开发和改进包括生成式AI在内的模型及安全措施的训练;用户可配置数据自动删除周期为3个月、18个月或36个月。
  • 覆盖服务生态:该政策不仅限于Google Search主应用,还延伸至Maps、Shopping、Flights、Hotels、Translate和News等所有搜索相关服务。

行业启示

  • 数据获取范式转移:行业正从被动爬取公开网络数据转向主动收集用户在封闭生态内产生的高价值、多模态交互数据,这对模型的真实世界表现力提升至关重要。
  • 隐私合规与用户体验的博弈:科技公司倾向于通过复杂的默认设置和细粒度选项来最大化数据留存率,用户需具备更高的数字素养以识别并调整隐私偏好。
  • 竞争态势加剧:随着Meta等竞争对手也在大规模利用用户媒体和硬件记录数据进行AI训练,数据独占性和质量将成为下一代AI服务竞争的核心壁垒。

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。

Gemini Gemini Training 训练 Policy 政策