Meta's three-month-old AI unit is already buckling from within
Drafted engineers, a monitoring petition and a livestream hijacked by an outburst: Meta's 6,500-strong AI push is fraying fast, and the tremors reach India.
The News
Meta's flagship applied artificial intelligence unit, barely three months old, is in open turmoil. According to reporting carried by TechCrunch, the roughly 6,500 engineers and product managers assigned to the group describe morale in freefall, with one calling the experience "literally the gulag" and another saying most people find the work "soul-crushing".
The friction spilled into view this week when an individual hijacked a livestreamed staff presentation with an expletive-laden outburst. Separately, more than 1,600 employees across the company signed a petition opposing a programme that monitors keystrokes and clicks. Chief executive Mark Zuckerberg addressed the unrest in an internal memo on Friday, conceding the recent changes had "caused distress".
The unit is run by Maher Saba, a 12-year Meta veteran who previously served as a vice-president in Reality Labs and now reports to chief technology officer Andrew Bosworth. Sitting above the effort is chief AI officer Alexandr Wang, who joined after Meta acquired his data-labelling firm Scale AI for $14.3 billion. Much of the assigned work involves manufacturing puzzles and coding problems used to train models, a task many engineers say feels beneath their training.
Why It Matters
The episode is a reminder that capital and headcount cannot manufacture culture. Meta has spent billions assembling a war chest of talent and compute, yet the people meant to deliver its AI ambitions describe being conscripted rather than recruited, with the choice framed as join or quit.
The pattern echoes the post-2022 reshuffles across Big Tech, when sudden reorganisations and mandatory relocations bred quiet resignation. The difference now is the stakes: this is the unit on which Zuckerberg has staked Meta's next decade. When Google folded teams into a single DeepMind structure in 2023, integration pains were real but the mission stayed legible to staff. Meta's problem looks sharper, because the headline work, generating training data, reads to many as a demotion dressed up as a priority.
There is also a governance signal here. A petition against keystroke monitoring, signed by 1,600 people, tells investors that the workforce powering the model roadmap does not trust how it is being managed. That is a retention risk hiding behind an impressive org chart.
Indian Angle
For India, the most direct read is on talent. Meta runs sizeable engineering operations in the country, and Indian engineers are heavily represented across global AI teams. A unit where senior staff feel demoted and surveilled becomes a hiring opportunity for Indian-headquartered model builders such as Sarvam and Krutrim, as well as the global capability centres in Bengaluru and Hyderabad now competing for the same profiles.
The story also throws light on the data-labelling economy that India helped build. The very work being described as soul-crushing, crafting puzzles and coding problems to train models, is the high-skill end of a pipeline that has long routed annotation and reinforcement tasks through Indian and South-East Asian vendors. Scale AI, now inside Meta, grew partly on that outsourced supply chain. Indian firms in this space should note that even premium in-house versions of this work are proving hard to staff happily.
Finally, there is a regulatory thread for MeitY and India's privacy watchdogs to watch. Keystroke and click monitoring of employees sits awkwardly beside the Digital Personal Data Protection framework. As Indian arms of global majors adopt similar productivity tooling, expect questions about employee consent and proportionality to surface here too.
FAQ
How large is Meta's troubled AI unit?
The applied AI group holds roughly 6,500 engineers and product managers. It was formed about three months ago and is led by Maher Saba, reporting to chief technology officer Andrew Bosworth, with Alexandr Wang as chief AI officer.
What triggered the petition?
More than 1,600 Meta employees signed a petition opposing a programme that monitors keystrokes and clicks. The backlash, alongside complaints about being drafted into the unit, prompted Mark Zuckerberg to acknowledge the distress in an internal memo on Friday.
What does this mean for Indian engineers?
Unrest at a flagship US AI team often loosens talent for rivals. Indian model builders like Sarvam and Krutrim, plus Bengaluru and Hyderabad capability centres, may find it easier to attract experienced engineers weighing their options.
Where can I read the original coverage?
The full account, drawing on internal communications, was reported by TechCrunch. The link appears in the attribution below.
This story was reported by TechCrunch. Read the full original coverage at TechCrunch.
Sources & Citations
Frequently Asked Questions
How large is Meta's troubled AI unit?
The applied AI group holds roughly 6,500 engineers and product managers. It was formed about three months ago and is led by Maher Saba, reporting to chief technology officer Andrew Bosworth, with Alexandr Wang as chief AI officer.
What triggered the petition?
More than 1,600 Meta employees signed a petition opposing a programme that monitors keystrokes and clicks. The backlash, alongside complaints about being drafted into the unit, prompted Mark Zuckerberg to acknowledge the distress in an internal memo on Friday.
What does this mean for Indian engineers?
Unrest at a flagship US AI team often loosens talent for rivals. Indian model builders like Sarvam and Krutrim, plus Bengaluru and Hyderabad capability centres, may find it easier to attract experienced engineers weighing their options.
Where can I read the original coverage?
The full account, drawing on internal communications, was reported by TechCrunch. The link appears in the attribution paragraph at the foot of this article.