Meta将追踪员工点击与按键数据以训练人工智能模型Meta to Track Employee Clicks and Keystrokes to Train AI Models

根据披露的信息，MCI软件将专门针对工作相关的应用程序和网站运行。Meta计划通过这种方式收集高精度的行为数据，以便其AI模型能够模仿人类在处理行政、编程或设计任务时的细微差别。这种监控不仅限于简单的文本输入，还包括了对用户界面交互的深度记录。

According to disclosed information, the MCI software will specifically run on work-related applications and websites. Meta plans to collect high-precision behavioral data this way, enabling its AI models to mimic the nuances of human behavior when handling administrative, programming, or design tasks. This monitoring extends beyond simple text input to include deep logging of user interface interactions.

• 监控对象：目前主要针对Meta在美国境内的员工。
• 数据类型：鼠标移动轨迹、点击频率、按键记录及屏幕截图。
• 运行环境：Meta配发的计算机、内部应用程序及工作相关网站。

• Target Audience: Currently primarily focused on Meta's U.S.-based employees.
• Data Types: Mouse movement trajectories, click frequency, keystroke logs, and screen snapshots.
• Operating Environment: Meta-issued computers, internal applications, and work-related websites.

Meta公司辩称，这种程度的数据采集对于开发下一代“AI智能体”至关重要。公司认为，传统的静态数据集无法提供足够的上下文信息来教导AI如何应对动态的办公环境。通过观察员工如何切换窗口、纠正输入错误以及在不同应用间导航，Meta旨在使其AI模型具备更强的逻辑推理和任务执行能力。

Meta argues that this level of data collection is essential for developing the next generation of "AI agents." The company believes that traditional static datasets cannot provide enough contextual information to teach AI how to navigate dynamic office environments. By observing how employees switch windows, correct input errors, and navigate between different applications, Meta aims to equip its AI models with stronger logical reasoning and task execution capabilities.

尽管Meta强调该工具主要用于训练AI，但这种全方位的监控不可避免地引发了关于员工隐私和职场压力的讨论。内部备忘录显示，该软件会捕获屏幕快照，这意味着员工在处理敏感信息或进行私人沟通时（即使是在工作设备上）也可能被记录。这种“按键级”的追踪在科技行业虽然并非首创，但Meta将其直接与AI训练挂钩的做法标志着企业数据利用的新阶段。

While Meta emphasizes that the tool is primarily for AI training, this all-encompassing monitoring inevitably sparks discussions about employee privacy and workplace pressure. Internal memos show the software captures screen snapshots, meaning employees' handling of sensitive information or private communications—even on work devices—could be recorded. While "keystroke-level" tracking is not a first in the tech industry, Meta's direct link between such tracking and AI training marks a new stage in corporate data utilization.

目前尚不清楚员工是否有权选择退出该计划，或者Meta将如何处理和匿名化这些高度个人化的行为数据。随着AI竞争进入白热化，Meta此举可能为其他科技公司开创先例，将员工的每一项数字资产都转化为公司的研发资源。

It remains unclear whether employees have the right to opt out of the program or how Meta will process and anonymize this highly personalized behavioral data. As the AI competition intensifies, Meta's move could set a precedent for other tech companies to convert every digital asset of their employees into corporate R&D resources.