Observed Signal · Apr 23, 2026 · Technical Release · Source: CNBC Technology · Impact: 4/5 · Sentiment: Negative
Meta tracks employee keystrokes for AI training
Meta has deployed an internal tool, the Model Capability Initiative (MCI), to capture employee on-screen inputs — including keystrokes, mouse clicks and navigation — across hundreds of websites and apps to build data for training its AI models. CNBC reviewed an internal memo and reporting says sites monitored include Google, LinkedIn, Wikipedia, GitHub, Slack, Salesforce, Atlassian and Meta properties such as Threads and Manus. The project is linked to CEO Mark Zuckerberg’s push to accelerate generative AI, including hires and new models like Muse Spark; Meta says safeguards are in place and the data will not be used for other purposes. Internal messages show employees raised privacy and sensitive-data exposure concerns. Reuters previously reported on the tracking tool.
A major platform (Meta) has launched an internal technical program to capture employee device interactions for AI training — this has broad implications for data governance, privacy, model training practices and potential regulatory scrutiny across the tech and ad ecosystem.
Track Meta Signals & Market Shifts in Real-Time
Polaris7 autonomous intelligence agents track regulatory filings, primary sources, executive changes, and deal flow 24/7. Create your free Explorer workspace to monitor these entities.
Key Takeaways & Evidence Grounding
- Meta created an internal tracking tool called Model Capability Initiative (MCI) to observe employee interactions on their work computers.
- MCI is configured to capture keystrokes, mouse clicks and on‑screen content on hundreds of third‑party sites and apps, including Google, LinkedIn, Wikipedia, GitHub, Slack, Salesforce and Atlassian, and Meta properties Threads and Manus.
- CNBC viewed an internal memo circulated by Meta Superintelligence Labs (MSL) which sought to explain safeguards; employees raised concerns about privacy and potential exposure of sensitive data.
- The project is tied to Meta CEO Mark Zuckerberg’s push to catch up in generative AI and follows hires such as bringing in Scale AI’s Alexandr Wang; Meta recently unveiled the Muse Spark model.
- A Meta spokesperson confirmed the project and said safeguards exist, that captured data would help train models, and that the data is not used for other purposes.
Connected Companies & Entities
5 Entities mappedOntology Mapping & Concepts
Related Market Signals & Shifts
Recent verified developments and strategic activity across this market segment.
Meta to Record Employee Keystrokes for AI Training
Meta plans to capture internal employee inputs — including mouse movements and keystrokes — via a newly launched internal tool to help train its AI models. Reuters first reported the initiative; TechCrunch obtained a Meta statement saying the tool will record interactions on certain applications, that safeguards exist to protect sensitive content, and that the data is used solely for model training. The move highlights growing industry practices of sourcing training data from internal corporate activity and raises privacy concerns about how historic internal communications and employee behavior may be harvested for AI development.
Meta Logs Employee Inputs for AI Training
Meta is rolling out a program called the Model Capability Initiative (MCI) that will run on employee work computers to capture mouse movements, keystrokes and intermittent screenshots across work apps and websites to generate training data for AI models. Reuters reported the initiative based on internal memos; a Meta spokesperson confirmed the tool and said data will not be used for performance evaluation and that security rules are in place to exclude "sensitive content," without detailing exclusions. Some employees described the monitoring as dystopian, and legal experts told Reuters that similar surveillance likely would be unlawful in Europe. Meta says the data will help make AI agents interact more naturally with computer interfaces. The deployment appears focused on the U.S. and comes amid broader Meta AI investments and workforce reductions concerns.
Meta Pauses Internal Employee Surveillance Program
Meta has suspended an internal tracking program introduced in April — called the Model Capability Initiative (MCI) — which recorded employees' mouse movements, clicks and keystrokes to help train its AI. Reuters, Business Insider and Wired reported that confidential employee data appeared to be accessible across Meta; Reuters reviewed internal documents indicating the data set included full prompts, transcripts, private conversations, personnel and performance data, and DSS confidentiality levels 1–4. Meta spokesperson Tracy Clayton said the company designed the program with privacy safeguards, that there is currently no evidence of unauthorized access, and that the initiative is being paused while an investigation is carried out. Reuters previously reported some collected information was stored unencrypted, raising employee privacy concerns.
Track Real-Time Market Signals & Shifts
Set up custom watchlists to receive automated, evidence-grounded executive digests whenever material signals or shifts occur across your tracked landscape.