Breaking

Meta Caps Employee AI Token Use, Pushes Staff Toward In-House Tools

June 20, 2026 at 17:00 EDT

Sales & Business
AI Agents
Foundation Models

Meta is reversing course on encouraging maximum AI use, imposing caps on internal AI token consumption and steering employees away from third-party tools such as Anthropic's Claude toward its own in-house tools. The company notified roughly 6,000 employees via an internal memo, against a backdrop of internal AI costs projected to reach billions of dollars in 2026.

Meta · Internal AI Policy Reversal

From "Tokenmaxxing" to Token Caps

After internal AI costs were projected to hit billions of dollars in 2026, Meta is capping token usage for about 6,000 employees and steering them away from external tools toward in-house ones.

~6,000

Employees facing new token limits

60–73.7T

Tokens consumed in ~30 days

2027

Formal per-team & per-person quotas begin

The cost of one month of "tokenmaxxing"

Public-pricing estimate of ~30 days of internal usage — a range spanning an order of magnitude.

~$900M

Low estimate

~$9B

High estimate

Up to ~10× more than the low end — enough to project billions in 2026 spend.

How the new controls roll out

AI GATEWAY

Real-time usage & spend tracking with anomaly alerts

→

FROM 2027

Per-team & per-person budgets and formal quotas

→

SHIFT IN TOOLS

Curb external tools; push in-house MetaMate

Cost-aware "tokenminimizing"

Supporters welcome the focus on value over raw volume — as leaders note, token count alone is not a measure of impact.

Gamification gone wrong

Leaderboards pushed workers to run parallel agents and repeat meaningless prompts to climb ranks — and caps may blunt adoption.

Not just Meta

Uber, Amazon and Walmart are also scrapping leaderboards or setting monthly caps. Uber reportedly burned through its 2026 budget in four months and now enforces a $1,500 per-tool monthly cap.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.