Meta has reportedly moved 30% to 50% of engineers off core and product engineering teams into new units focused on data labeling and RLHF for AI training. The reassignments are said to have begun in late April 2026, with large-scale notifications going out through May.
May 2026 · Meta
Meta Pulls Engineers Off Core Teams to Label AI Training Data
After late April 2026, Meta began reassigning engineers from ads, infrastructure and security into data-labeling work for AI agents — framed first as "volunteer," later as mandatory transfers. On some core teams, 30–50% of engineers are said to be affected.
1,000+
Initial transfers into the Applied AI org (~April 2026)
~6,500
In the Agent Data Optimization group (4,000–5,000 engineers)
30–50%
Of engineers shifted on some core infra & security teams
Who ends up doing data work
Estimated share of Meta's ~25,000 engineers — roughly 1 in 5 to 1 in 6
The loop being built
Engineers
→
Training data
→
AI agents
The backdrop
Seen internally as a push to catch OpenAI and Google in AI agents — by putting top talent directly on data generation.
Up to $135B
Planned AI infrastructure spend, 2026
~$14.3B
For ~49% of Scale AI (2025)
~8,000
Layoffs (~10%) coinciding with the shift
Seen positively
Faster AI tooling and adoption
New "AI-native pods" structure
In-house talent over outside contractors
The pushback
$500K engineers on labeling seen as inefficient
Building data that may replace their own jobs
Petition vs. MCI drew 500+ signatures; union talk
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…