Meta Layoffs: Inside Meta's 'AI Draft' As 7,000 Employees Reassigned to New AI Teams After 8,000 Workers Laid Off
Meta laid off roughly 8,000 workers while reassigning 7,000 to core AI teams like Applied AI under CTO Andrew Bosworth. Staff labelled the mandatory shift an 'AI draft', with roles focusing on data labelling. Meta is also tracking workflows via its Model Capability Initiative to train models on real-world examples of human-computer tasks.
Meta has initiated a sweeping corporate overhaul that places artificial intelligence at the center of its long-term strategy, cutting thousands of traditional roles while aggressively reassigning existing personnel. The social media giant has laid off approximately 8,000 employees globally, while simultaneously reallocating roughly 7,000 workers to newly formed AI divisions. According to internal memos reviewed by Business Insider, affected personnel were informed via email that they had been selected to join a newly established Applied AI (AAI) group and other specialised model-development teams.
Internal 'AI Draft' Deployed by Meta
The sudden restructuring has drawn mixed reactions across Meta's workforce, with employees using internal communication channels and private Discord servers to discuss the abrupt shifts. Because participation in the reassignment initiative is mandatory, several workers have informally labelled the transition as an "AI draft". Meta Layoffs: Former Engineer Shares Lessons on Surviving Tech Job Cuts, Importance of Personal Branding and Financial Planning.
"I got drafted," one employee posted on an internal forum. Another worker quickly responded, "Welcome to the draft." While some staff members expressed relief at surviving the global layoffs, many noted widespread ambiguity regarding their new day-to-day responsibilities. Multiple employees indicated they believe these newly formed task forces will focus heavily on intensive data labelling and model training, which includes manual tasks such as tagging large image datasets and correcting chatbot responses to refine system accuracy.
The Structure of Applied AI at Meta
The reallocated workforce will operate under a heavily flattened management hierarchy. The newly created Applied AI organisation is led by Engineering Vice President Maher Saba and reports directly to Chief Technology Officer Andrew Bosworth. According to internal documentation, the restructuring divides the workforce into distinct operational units, including the "Agent Transformation Accelerator" and the "Agent Data and Optimization" groups. Rather than focusing on theoretical science, these teams are specifically tasked with accelerating real-world AI agent integration and refining foundational model training pipelines. To support this pivot, Meta has begun shifting internal employee job titles to "AI builder" and reorganising staff into compact, project-focused AI units.
Data Engines and Employee Monitoring by Meta
The shift toward a labour-intensive data engine aligns closely with the background of Meta's Chief AI Officer, Alexandr Wang, who previously founded and managed the data-labelling firm Scale AI before joining the technology conglomerate. In a leaked audio recording of an internal meeting, CEO Mark Zuckerberg highlighted Wang's specialised background, suggesting that Meta could secure a distinct competitive advantage by mobilising its own massive internal workforce to train and optimise proprietary models. Meta Layoffs 2026: What H-1B Visa Employees Were Told After Job Cuts Affecting Nearly 8,000 Workers.
"I think that this is going to be a very big advantage if we can do it," Zuckerberg stated during the meeting, though he noted at the time that the strategy remained a "hypothesis". To support this model-training push, Meta deployed an internal tracking software program known as the Model Capability Initiative. The software systematically monitors employee computer interactions, capturing keystrokes, screenshots, and workflows to provide real-world human examples for AI training. An internal post announcing the deployment clarified the project's parameters, stating, "For agents to understand how people actually complete everyday tasks using computers, we need to train our models on real examples."
(The above story first appeared on LatestLY on Jun 02, 2026 05:53 PM IST. For more news and updates on politics, world, sports, entertainment and lifestyle, log on to our website latestly.com).
