Meta Platforms, Inc. has paused its internal artificial intelligence training program, the Model Capability Initiative (MCI), following concerns that sensitive employee information was accessible to a wider audience than intended. The tool, launched in April, recorded mouse movements, clicks, and keystrokes on work computers to train Meta's AI models using data from U.S.-based employees. The company confirmed the temporary halt while it investigates potential data security issues, though a spokesperson stated there is currently no indication that data was improperly accessed.

The decision to pause the initiative followed the filing of a high-priority internal security incident report. Internal documents indicated that the exposed information included AI prompts, transcripts, private conversations, employee performance-related information, and internal sensitivity classifications. Despite the announced pause, a source noted that the system was still collecting information on Monday afternoon, with the company acknowledging that a complete shutdown would take time.

Recent reports had already raised questions about the program, suggesting it was gathering more information than originally described and storing some data without encryption. Employees involved in internal discussions expressed concern that personal tax and medical information accessed on work computers was at risk. Staff had been assured that data would be protected and used only for legitimate business purposes following "aggressive filtering."

Key Details of the Paused Initiative

Stock Performance

Shares of Meta closed Monday at $563.85, down 2.32%. The stock slipped an additional 0.20% to $562.70 in after-hours trading.

Meta Platforms stock is underperforming the market this year, ending the week at $577, down by 27% from its highest point last year. Its valuation has fallen from $2 trillion to $1.47 trillion. The decline is attributed to ongoing investments in artificial intelligence and fears of shareholder dilution, as the company plans to spend $135 billion this year, up from previous guidance of $115 billion. Last year, it spent $72 billion on capex.

Valuation metrics drop

The stock retreat has pushed Meta's trailing price-to-earnings ratio to 20, its lowest level since February 27 2023. The ratio peaked at 32 in October last year. The forward price-to-earnings ratio has slipped to 17, below the technology sector average of 33 and its five-year average of 22. The forward price-to-earnings-to-growth ratio has dropped to 86 cents, compared to the sector median of 1.2.

Funding and AI competition

Meta is considering raising about $85 billion to fund capital investments, following Alphabet's move to raise $80 billion through debt and equity. Concerns persist regarding the return on investment for AI spending. Meta AI holds a smaller market share compared to competitors like ChatGPT, Claude, and Google Gemini. A recent report indicated that Grok, Perplexity, DeepSeek, and Meta AI command less than 5% market share.

Analysts remain bullish

Despite the decline, analysts maintain a bullish outlook. The average estimate among analysts is $834, up by 45% from the current level. RBC Capital maintained its target at $810, while Rosenblatt kept its target at $1,015. Wedbush boosted its rating from neutral to outperform, and Mizuho hiked its target to $850.

The most recent earnings showed revenue jumped by 33% to $56.3 billion, while net income soared by 61% to $26 billion. Analysts estimate revenue will jump by 25% this year to $237 billion and exceed $282 billion next year.

Technical risks

Technical indicators suggest further downside risk. The weekly chart shows the stock has slumped below the 23.6% Fibonacci Retracement and the 50-week moving average. A multi-month head-and-shoulders pattern, a bearish reversal sign, has formed. The MACD lines continue to fall, suggesting the stock may drop to $500.