Meta has quietly begun installing an internal tracking tool on devices used by its U.S. employees that captures mouse movements, keystrokes and periodic screenshots to train artificial intelligence systems — part of a broader reorganization that puts AI at the center of how work gets done inside the company.
What’s being rolled out
- The tool, dubbed the Model Capability Initiative (MCI), runs across work applications and websites and records routine interactions — keyboard shortcuts, menu navigation and other common workflows. An internal memo from Meta’s SuperIntelligence Labs confirms the system can capture periodic screenshots.
- Company documents say the data collected is intended to teach models how people actually use computers, letting AI learn from day-to-day behavior “simply by doing their daily work.”
Why Meta is doing this
- Meta is shifting internal operations toward AI-driven workflows. Employees are being encouraged to rely more on AI tools for tasks such as coding, even if that temporarily slows productivity during the transition.
- The company has created a new “AI builder” role, reshaped engineering teams, and formed an Applied AI engineering group whose brief is to build systems that can write and deploy software. Some engineers have already been reassigned to these efforts.
Leadership messaging and data use
- CTO Andrew Bosworth said Meta is expanding data collection under its “AI for Work” program, now renamed the Agent Transformation Accelerator. “The vision we are building towards is one where our agents primarily do the work and our role is to direct, review and help them improve,” Bosworth wrote, adding that the systems are designed to flag moments where human input is needed.
- Bosworth also said Meta is focused on building datasets and evaluation frameworks based on employee workflows, with workplace interaction data central to development.
- Meta spokesman Andy Stone stressed MCI data is “used only to train models and not to assess employee performance,” adding that models need real examples of clicking buttons and navigating menus to be effective.
- The company says safeguards exist to avoid collecting sensitive data, though it hasn’t specified what is excluded.
Beyond workflows: a digital CEO and shopping integrations
- Meta is also experimenting with a photorealistic digital version of CEO Mark Zuckerberg trained on his voice, tone and communication style to converse with employees in real time — a project intended to give staff more direct access to leadership across a global workforce. The digital avatar is still in early development.
- Separately, Meta continues to tighten the connection between creators, AI tools, advertising and shopping on Instagram and Reels, aiming to shorten the path from discovery to purchase.
Why this matters to crypto and web3 audiences
- Data provenance and model training at this scale raise familiar privacy and governance questions that web3 projects often aim to address with decentralized identity, verifiable consent, and on-chain auditability.
- The move underscores how major platforms are building AI-driven creator and commerce experiences — a space where tokenized creator economies, programmable royalties and decentralized marketplaces could compete or interoperate with centralized offerings.
- For builders and investors, Meta’s push signals accelerating demand for workflow-tracking datasets, agent orchestration tools and real-time digital avatars — areas that may spawn new tooling, standards and opportunities for integration with blockchain-based identity and payments.
Bottom line: Meta is intensifying its internal data collection to accelerate AI agents that can do routine work, while also investing in digital avatars and smoother commerce experiences. The effort raises practical privacy questions and highlights opportunities — and competition — for web3 approaches to identity, consent and creator monetization.
Read more AI-generated news on: undefined/news
