Meta’s AI overhaul rattles engineering teams

Meta’s engineering organization has shifted sharply from a culture built around autonomy, fast shipping and stable infrastructure toward a top-down AI mandate. The company’s push intensified after Meta acquired a 49% stake in Scale AI for ?.8B and brought in Scale AI CEO Alexandr Wang to lead AI strategy, with an emphasis on training data, labeling and RLHF for coding models.

Engineers have been enrolled in systems that track keystrokes and mouse clicks to generate training data, initially with no opt-out. Product, infrastructure and security teams have also been directed to reassign 30-50% of engineers into Agent Data Optimisation work, where many perform repetitive data-labeling and model-feedback tasks. The unit is described as having around 6,500 people, including roughly four to five thousand software engineers, while Meta’s broader engineering workforce is around 25,000.

The changes have collided with layoffs, aggressive performance reviews and AI token-usage metrics, creating incentives for engineers to maximize AI use rather than focus on core product work. Internal accounts link depleted security staffing and AI-reviewed code to a serious Instagram account takeover incident on 30 May, followed by CISO Guy Rosen’s departure after the issue was resolved on 1 June. Frustration inside Meta has escalated as senior leaders acknowledge a brutal environment and engineers consider leaving.