Meta pauses an AI training program that tracked employee keystrokes after a configuration error exposed sensitive data companywide. Internal backlash grows as the security review unfolds.
Alpha Score of 57 reflects moderate overall profile with poor momentum, strong value, strong quality, weak sentiment.
Meta is pausing an internal AI training program that tracked employee keystrokes after sensitive data from the project was accidentally exposed companywide, according to internal screenshots obtained by Business Insider.
The program collected keystroke data from a subset of workers to train Meta's AI models. A configuration error made the training data accessible to all employees, not just the research team. The exposed data included keystroke logs and metadata that could identify individual participants, raising privacy and security concerns inside the company.
The pause comes as Meta faces growing scrutiny over its data-handling practices, both internally and from regulators. Employees have pushed back against the program since its launch, arguing that keystroke monitoring blurs the line between productivity tracking and surveillance. The leak has amplified those concerns, with some staff calling for a permanent halt and a third-party audit of the program's data governance.
Meta has not commented publicly on the incident. The company's internal security team is reviewing the configuration error and assessing whether any of the exposed data was accessed by unauthorized employees. The program's future will depend on that review, according to people familiar with the matter.
The incident adds to a string of data-management missteps at Meta. The company has faced fines and regulatory actions in Europe over GDPR violations. Its use of employee data for AI training has been a recurring flashpoint in internal debates about workplace privacy.
For Meta, the immediate risk is operational: the pause delays the AI training pipeline that the program was feeding. The larger question is whether the company can rebuild trust with its workforce on data handling. The leak suggests that internal controls have not kept pace with the scale of Meta's AI ambitions.
AlphaScala's proprietary score for Meta (META) stands at 57 out of 100, labeled Moderate, reflecting balanced risk-reward at the current price of $563.73, down 2.34% on the session. The stock has been under pressure this year as the market weighs the costs of Meta's AI infrastructure buildout against potential revenue gains.
The internal review is expected to conclude within two weeks. No decision on the program's restart has been announced.
Prepared with AlphaScala research tooling and grounded in primary market data: live prices, fundamentals, SEC filings, hedge-fund holdings, and insider activity. Each story is checked against AlphaScala publishing rules before release. Educational coverage, not personalized advice.